Identification of human zonulin, a physiological
modulator of tight junctions, as prehaptoglobin-2
Amit Tripathia, Karen M. Lammersa, Simeon Goldbluma, Terez Shea-Donohuea, Sarah Netzel-Arnettb,
Marguerite S. Buzzab, Toni M. Antalisa,b, Stefanie N. Vogela,c, Aiping Zhaoa, Shiqi Yanga, Marie-Claire Arriettac,
Jon B. Meddingsd, and Alessio Fasanoa,1
aMucosal Biology Research Center,bCenter for Vascular and Inflammatory Diseases and Department of Physiology, andcDepartment of Microbiology
and Immunology, University of Maryland School of Medicine, Baltimore, MD 20201; anddDepartment of Medicine, University of Alberta, Edmonton,
Alberta, Canada T6G 2V2
Communicated by Maria Iandolo New, Mount Sinai School of Medicine, New York, NY, June 25, 2009 (received for review December 16, 2008)
Increased intestinal permeability (IP) has emerged recently as a
common underlying mechanism in the pathogenesis of allergic, in-
lin, the only physiological mediator known to regulate IP reversibly,
has remained elusive. Through proteomic analysis of human sera, we
have now identified human zonulin as the precursor for haptoglo-
bin-2 (pre-HP2). Although mature HP is known to scavenge free
hemoglobin (Hb) to inhibit its oxidative activity, no function has ever
been ascribed to its uncleaved precursor form. We found that the
single-chain zonulin contains an EGF-like motif that leads to transac-
tivation of EGF receptor (EGFR) via proteinase-activated receptor 2
(PAR2) activation. Activation of these 2 receptors was coupled to
increased IP. The siRNA-induced silencing of PAR2 or the use of
PAR2?/?mice prevented loss of barrier integrity. Proteolytic cleavage
of zonulin into its ?2- and ?-subunits neutralized its ability to both
activate EGFR and increase IP. Quantitative gene expression revealed
celiac disease. To our knowledge, this is the initial example of a
molecule that exerts a biological activity in its precursor form that is
distinct from the function of its mature form. Our results therefore
characterize zonulin as a previously undescribed ligand that engages
a key signalosome involved in the pathogenesis of human immune-
autoimmune diseases ? epidermal growth factor receptor ?
gut permeability ? proteinase-activated receptor 2 ? celiac disease
demic’’ of allergic, inflammatory, and autoimmune diseases
recorded in industrialized countries during the past 3–4 decades
(1). Apart from genetic makeup and exposure to environmental
triggers, a third key element [i.e., increased intestinal perme-
ability (IP)] has been proposed in the pathogenesis of these
diseases (2–4). IP, together with antigen sampling by enterocytes
and luminal dendritic cells, regulates molecular trafficking be-
tween the intestinal lumen and the submucosa, leading to either
tolerance or immunity to non–self-antigens (5). However, the
dimensions of the paracellular space (10–15 Å) suggest that
solutes with a molecular radius exceeding 15 Å (?3.5 kDa)
(including proteins) are normally excluded from this uptake
route. The intercellular tight junctions (TJs) tightly regulate this
paracellular antigen trafficking. TJs are now appreciated to be
extremely dynamic structures operative in several key functions
of the intestinal epithelium under both physiological and patho-
logical circumstances (3). However, despite major progress in
our knowledge regarding the composition and function of in-
tercellular TJs, the mechanism(s) by which they are regulated
is(are) still incompletely understood. The discovery of Vibrio
cholerae zonula occludens toxin (Zot), a toxin that increases TJ
permeability, led us to the identification of its eukaryotic
counterpart, zonulin, as the only physiological mediator known
to regulate IP reversibly by modulating intercellular TJs (6, 7).
ncreased hygiene leading to a reduced exposure to various
microorganisms has been implicated as a cause for the ‘‘epi-
Human zonulin is a ?47-kDa protein that increases IP in
nonhuman primate intestinal epithelia (7), participates in intes-
tinal innate immunity (8), and is overexpressed in autoimmune
disorders in which TJ dysfunction is central, including celiac
disease (CD) (9, 10) and type 1 diabetes (T1D) (11). Although
zonulin’s role as an intestinal permeating modulator in health
and disease has been described functionally, its biochemical
characterization has remained elusive. Through proteomic anal-
ysis of human sera, we report herein that zonulin is identical to
the precursor of haptoglobin-2 (pre-HP2), a molecule that, to
date, has only been regarded as the inactive precursor for HP2,
one of the two genetic variants (together with HP1) of human
HPs (see Fig. S1). Our studies demonstrate the previously
undescribed functional characterization of zonulin as pre-HP2,
a multifunctional protein that, in its intact single-chain precursor
form, appears to regulate IP by transactivating the epidermal
growth factor receptor (EGFR) via proteinase-activating recep-
tor 2 (PAR2) activation, whereas in its cleaved 2-chain form, it
acts as an Hb scavenger.
Characterization of Zonulin from CD Human Sera. Because zonulin is
detected in human sera by a zonulin cross-reacting anti-Zot
Ab-based ELISA (7–10) and is increased in patients with CD
compared with normal controls (10), we initially used Western
blot (WB) analysis to detect zonulin immunoreactivity of pro-
sera displayed 2 major protein bands with apparent molecular
weights (MWs) of 18 and 9 kDa (Fig. 1). Three distinct patterns
of reactivity were identified in CD sera: an 18-kDa protein band
(Fig. 1, lane 1), a 9-kDa protein band (Fig. 1, lane 2), and both
9- and 18-kDa protein bands (Fig. 1, lane 3). Of note, a ?45-kDa
band was detected only in sera that displayed the single 18-kDa
band (Fig. 1, lane 1) but was not detected in sera with either the
9-kDa band or both bands (Fig. 1, lanes 2 and 3). Two-
who expressed the 18-kDa band revealed 2 zonulin immunore-
active spots [see supporting information (SI) Text and Fig. S1 A
and B] that were subjected to MS/MS analysis. The 18-kDa spot
was identified as the ?2-chain of HP2 (accession no. GI:223976)
and the 9-kDa spot as the ?1-chain of HP1 (accession no.
Author contributions: S.G., T.S.-D., T.M.A., S.N.V., M.-C.A., J.B.M., and A.F. designed re-
M.S.B., and A.F. contributed new reagents/analytic tools; A.T., K.M.L., S.G., T.S.-D., S.N.-A.,
M.S.B., T.M.A., S.N.V., A.Z., S.Y., M.-C.A., J.B.M., and A.F. analyzed data; and S.G., T.S.-D.,
T.M.A., S.N.V., J.B.M., and A.F. wrote the paper.
Conflict of interest statement: A.F. and S.N.V. have financial interest in Alba Therapeutics,
a company involved in the development of treatments of CD alternative to the GFD.
Freely available online through the PNAS open access option.
1To whom correspondence should be addressed. E-mail: email@example.com.
This article contains supporting information online at www.pnas.org/cgi/content/full/
September 29, 2009 ?
vol. 106 ?
no. 39 ?
GI:3337390). A diagram showing the structure of 2-chain HP1
and HP2 and their precursors is presented in Fig. S1C. A random
screening of 14 sera from CD patients revealed that 7% were
HP1 homozygous, 57% were HP1/HP2 heterozygous, and 36%
were HP2 homozygous (Fig. S1D).
Characterization of Zonulin from Human HP Preparations.Toconfirm
the identity of the immunoreactive bands recognized by the
polyclonal zonulin cross-reacting anti-Zot IgG Ab in human CD
sera, commercially purified preparations of human HP from
subjects homozygous for either HP1 (HP1–1) or HP2 (HP2–2)
were simultaneously resolved on a single gel by SDS/PAGE and
analyzed by Coomassie staining (Fig. 2A). As expected, the
?1-chain of HP1–1 exhibited a MW of ?9 kDa (Fig. 2A, lane 1),
whereas the ?2-chain of HP2–2 had a MW of ?18 kDa (Fig. 2A,
lane 2). Because of its glycosylation, the ?-chain exhibited a MW
of ?52 kDa in both HP1–1 and HP2–2 preparations (Fig. 2A,
lanes 1 and 2). After a 3-h deglycosylation reaction with N-
glycosidase F (PGNase F), the ?-chain of both HP1–1 and
HP2–2 ran as multiple bands below 52 kDa, presumably attrib-
utable to varying degrees of deglycosylation (Fig. 2A, lanes 3 and
4). As anticipated, after glycosidase treatment, no changes in gel
mobility for either the ?1-chain of HP1–1 (Fig. 2A, compare
lanes 1 and 3) or the ?2-chain of HP2–2 (Fig. 2A, compare lanes
2 and 4) were evident.
Fig. 2B presents immunoblots of commercially available pu-
rified homozygous HP1–1 and HP2–2 proteins both before and
after deglycosylation. Proteins were run simultaneously on a
single gel and immunoblotted with polyclonal zonulin cross-
reacting anti-Zot Ab (Fig. 2B Left), monoclonal antiglycosylated
?-chain HP (Fig. 2B Center), or polyclonal anti-HP Ab (Fig. 2B
Right). Anti-Zot Ab reacted strongly with both the HP1–1
?1-chain and the HP2–2 ?2-chain (Fig. 2B Left, lanes 1 and 2,
respectively) and revealed an additional band at ?45 kDa
present in the HP2–2 but not the HP1–1 preparations (Fig 2B
Left, arrows). As expected, the monoclonal anti-HP Ab, raised
against the ?52-kDa HP ?-glycosylated subunit, recognized only
the ?-chain of either HP1–1 or HP2–2 (Fig. 2B Center, lanes 1
and 2, respectively), whereas the polyclonal anti-HP Ab recog-
nized epitopes of the ?1-, ?2-, and ?-chains of both HP1–1 and
HP2–2 (Fig. 2B Right, lanes 1 and 2, respectively). Fig. 2B also
shows immunoblotted HP1–1 and HP2–2 preparations after
deglycosylation using the same 3 Ab. The pattern of reactivity of
the 3 Ab tested for the nonglycosylated 9-kDa ?1-subunit and the
18-kDa ?2-subunit did not change after deglycosylation (Fig. 2B,
lanes 3 and 4, respectively). However, deglycosylation caused the
expected gel mobility shift of the ?-chain in both HP1–1 and
HP2–2. The monoclonal anti-HP Ab (Fig. 2B Center, lanes 3 and
patient sera samples that were depleted of albumin and immunoglobulins.
Three main patterns were detected: sera showing an 18-kDa immunoreactive
band and a fainter ?45-kDa band (lane 1), sera showing only a 9-kDa band
(lane 2), and sera showing both the 18- and 9-kDa bands (lane 3).
at the predicted MWs of 9 and 18 kDa, respectively. Deglycosylation with PGNase caused a shift of the ?-chain to a MW of ?36 kDa (complete deglycosylation)
or higher (incomplete deglycosylation). As expected, no shifts were observed in the nonglycosylated ?1- and ?2-chains. (B) WB of purified human homozygote
using polyclonal anti-Zot (Left), monoclonal anti-HP (Center), or polyclonal anti-HP (Right) Ab. The polyclonal Ab tested recognized both the ?1- and ?2-chains
(lanes 1 and 2), whose pattern of reactivity did not change after deglycosylation of both HP1–1 and HP2–2 protein preparations (lanes 3 and 4). Conversely,
deglycosylation caused the expected gel mobility shift of the ?-chain in both HP1–1 and HP2–2 detected by either the anti-HP monoclonal (Center, lanes 3 and
4) or anti-HP polyclonal (Right, lanes 3 and 4) Ab. The zonulin cross-reacting anti-Zot Ab recognized an extra ?45-kDa band in HP2–2 but not in HP1–1 that did
not shift after deglycosylation (arrows). MS/MS analysis and N-terminal sequencing identified this ?47-kDa band as pre-HP2.
Coomassie and Western immunoblotting of purified human homozygote HP1–1 and HP2–2 both untreated and after deglycosylation with PGNase. (A)
www.pnas.org?cgi?doi?10.1073?pnas.0906773106Tripathi et al.
4) recognized only 2 incomplete deglycosylated ?-chain bands,
whereas the polyclonal anti-HP Ab also recognized the com-
pletely deglycosylated ?36-kDa ?-chain (Fig. 2B Right, lanes 3
and 4). The 45-kDa band that was present only in the HP2–2
preparation and recognized by anti-Zot Ab did not show any
change in gel mobility on deglycosylation, but it appeared less
intense (Fig. 2B Left, lane 4). MS/MS analysis and NH2-terminal
sequencing of this 45-kDa protein band performed on 2 distinct
samples analyzed at different times identified this protein as the
human pre-HP2 (accession no. P00738). The combined MS/MS
analyses covered a total of 49.8% of nonoverlapping protein and
13 unique peptides spanning the entire protein sequence. There-
the uncleaved single-chain pre-HP2 but not the ?-chain. These
results suggest that the anti-Zot Ab used to measure serum
zonulin by ELISA should supposedly detect the highly abundant
HP1 and HP2 proteins as well as pre-HP2. However, the amount
of serum zonulin detected by ELISA is in the ng/mL range (11),
whereas the entire HP pool in serum is in the mg/mL range (12).
To address this apparent discrepancy, we repeated the WB
analysis of both human sera and purified HPs under nondena-
turing conditions using anti-Zot Ab (see SI Text). The WB
showed a series of bands in HP2–2 phenotype sera (Fig. S2A)
and in commercially purified HP2–2 (Fig. S2B), although no
bands were detected in either HP1–1 phenotype sera (Fig. S2A)
or in commercial purified HP1–1 (Fig. S2B). Conversely, the
anti-HP polyclonal Ab, which did not recognize the uncleaved
pre-HP2 (Fig. 2B), detected bands in both commercially purified
HP1–1 and HP2–2 preparations. Combined, these data suggest
that under nondenaturing conditions, the anti-Zot Ab detect
only the single-chain pre-HP2 but not the 2-chain mature HPs,
further supporting the notion that the single-chain pre-HP2, but
not its cleaved 2-chain mature form, corresponds to the zonulin
Functional Analysis of Recombinant Zonulin. The primary transla-
tion product of the mammalian HP2 mRNA transcript is a
polypeptide that dimerizes cotranslationally and is proteolyti-
cally cleaved while still in the endoplasmic reticulum by the
serine Cr1-like protease (Cr1LP) (13). Conversely, zonulin is
detectable in human serum as uncleaved pre-HP2 (Fig. 2 and
Fig. S2). To confirm the identification of zonulin as the single-
chain pre-HP2 and not the cleaved mature 2-chain HP2, we
expressed recombinant pre-HP2 by inserting the pre-HP2 cDNA
into an insect cell vector and expressed it using a baculovirus
expression system. We obtained highly purified recombinant
pre-HP2 that was recognized by the anti-Zot polyclonal Ab
similar to Fig. 2B and that migrated at an apparent MW of ?53
kDa because of the 6xHis tag attached at the C-terminus (Fig.
S3A). The single-chain pre-HP2 was then subjected to proteo-
lytic cleavage using a series of serine proteases. Matriptase,
urokinase, thrombin, and plasma kallikrein did not cleave pre-
HP2, whereas plasmin caused complete degradation of the
protein (Fig. S3B). In contrast, treatment with the intestinal
serine protease trypsin led to the appearance of 2 major bands
that migrated with MWs compatible with the ?2- and ?-subunits
of zonulin (Fig. S3B). NH2-terminal sequencing of these 2 bands
showed the 2 proteins to be identical to the pre-H2 ?2- and
?-chains cleaved at the predicted161Arg cleavage site. The intact
single-chain pre-HP2 and the cleaved 2-chain mature HP2
obtained after trypsin digestion were both tested for their
biological activities in the studies below.
Ex Vivo Effect of Recombinant Zonulin on TEER in Mouse Small
Intestine Mounted in the Microsnapwell System. Recombinant pre-
HP2 (henceforth defined as zonulin) was applied to WT
C57BL/6 murine small intestine segments mounted in micros-
napwells. Recombinant single-chain zonulin added to the mu-
cosal (luminal) aspect of mouse intestinal segments decreased
transepithelial electrical resistance (TEER) (i.e., increased per-
meability) when applied at concentrations ?40 ?g/mL (Fig. 3).
trypsin-cleaved 2-chain HP2 was tested (Fig. 3).
In Vivo Effect of Recombinant Zonulin on Mouse Gastrointestinal
Permeability. To establish whether zonulin might alter IP in vivo,
mice were gavaged with zonulin (170 ?g per mouse), and
gastroduodenal permeability and small intestine permeability
were tested using specific sugar probes (sucrose and lactulose/
small intestinal and gastroduodenal permeability compared with
BSA-treated controls (Table 1). Gastroduodenal permeability
and small intestine permeability each returned to baseline within
48 h following exposure to zonulin (Table 1).
To determine whether the 2-chain mature HP2 affected IP,
the in vivo experiments described previously were repeated by
dependent manner. Zonulin was applied to the luminal side of C57BL/6 WT
applied at a rate of 200 ?g/mL. Starting at 60 min postexposure, zonulin
(P value ranging from 0.03–0.036). Data are mean values ? SEM from 4
Zonulin increased IP in C57BL/6 WT mice in a dose- and time-
Table 1. Effect of zonulin on mouse gastroduodenal (sucrose) and small intestinal (lacman)
permeability in vivo
ChallengeRecovery After 48 h
68.44 ? 17.52*
8.75 ? 6.67
?0.40 ? 2.91
22.91 ? 5.4†
0.09 ? 4.40
0.03 ? 1.54
1.33 ? 3.59
0.60 ? 6.14
0.94 ? 2.19
0.48 ? 1.78
*Sucrose P ? 0.0049 compared with both BSA control and 2-chain HP2 (n ? 10 for each group of treatment).
†Lacman P ? 0.0024 compared with both BSA control and 2-chain HP2.
Tripathi et al.PNAS ?
September 29, 2009 ?
vol. 106 ?
no. 39 ?
to the single-chain zonulin, 2-chain HP2 (170 ?g per mouse)
failed to alter either gastroduodenal or small intestine perme-
ability compared with BSA-treated controls (Table 1). Com-
bined, these data indicate that the single-chain zonulin, but not
its 2-chain mature HP2 form generated by proteolytic cleavage,
retains the reversible permeating activity previously reported for
Transcriptional Expression of Zonulin in Human Duodenal Tissues.
Zonulin mRNA expression and quantification in human intestinal mucosae.
Using specific primers and the cDNA of human intestinal
biopsies from zonulin-positive subjects, we amplified a 686-bp
fragment, of which 144 bp belong to the ?-chain and 542 bp
belong to the ?-chain of both HP1 and HP2 genes. Sequencing
of this fragment confirmed its identity as HP, but HP1 could
not be distinguished from HP2 because of the common
sequence in the amplified region. To overcome this and
specifically to quantify the expression of the zonulin gene in the
human intestine, cDNA obtained from the intestinal mucosae
of healthy individuals (n ? 10), CD patients with acute-phase
disease (n ? 7), and CD patients with disease in remission
following a gluten-free diet (GFD) (n ? 3) was analyzed by
real-time PCR using primers and probes specific for the
?2-chain. Compared with healthy individuals, zonulin mRNA
expression was increased in the intestinal mucosae of CD
subjects with active disease (3-fold increase; P ? 0.05). Intes-
tinal mucosae of 3 CD subjects adhering to a GFD showed only
a 1.5-fold increase in zonulin expression compared with con-
trols (Fig. S4).
Recombinant Zonulin Activates EGFR and Causes TEER Changes Through
PAR2. It has recently been reported that gliadin, a glycoprotein
present in wheat and several other cereals and identified as the
environmental trigger responsible for the autoimmune damage
of the small intestine typical of CD (15), fully reproduces the
effects of EGF on the actin cytoskeleton (16), effects that are
very similar to those previously reported for zonulin (7, 10, 16).
Furthermore, structural analysis revealed that the pre-HP2
?-chain includes an EGF motif that contains 6 spatially
conserved cysteine residues that form 3 intramolecular disul-
fide bonds (Fig. S1C) necessary for EGF-like activity. To
determine whether zonulin can activate EGFR, increasing
concentrations of baculovirus-derived recombinant zonulin
were added to Caco-2 intestinal epithelial cells. The cells were
lysed, immunoprecipitated with anti-EGFR Ab, and processed
for phosphotyrosine immunoblotting (PY-Plus). At concen-
trations ?15 ?g/mL, zonulin increased tyrosine phosphoryla-
tion of EGFR (Fig. 4A and Fig. S5A). To establish the role of
EGFR in zonulin-induced alterations in TEER further, we also
performed both the in vitro and ex vivo experiments described
previously in the presence of the EGFR-selective protein
tyrosine kinase (PTK) inhibitor AG1478. Preincubation of
Caco-2 cells for 2 h with AG1478 (5 ?M) prevented zonulin-
induced EGFR phosphorylation on Y1068 (Fig. 4B and Fig.
S5B). Similarly, pretreatment with AG1478 abolished the
reduction in TEER in response to zonulin (Fig. 4C). Finally,
trypsin digestion of zonulin dramatically reduced its ability to
activate EGFR (Fig. 4D). Combined, these data suggest that
the single-chain zonulin activates EGFR and induces an
EGFR-driven decrease in TEER, whereas the cleaved 2-chain
HP2 fails both to activate EGFR and to increase IP.
Several G protein-coupled receptors, including PAR2 (17),
transactivate EGFR (18). Because Zot and zonulin share a
similar mechanism of action (6) and the zonulin protein se-
quence contains a Zot-like and PAR2-activating peptide (AP)–
like motif in its ?-chain (FCAGMS), we asked whether zonulin-
induced EGFR activation might be dependent on PAR2.
Experiments in Caco-2 in which PAR2was silenced and exper-
iments in PAR2?/?mice demonstrated that zonulin induced
PAR2-dependent transactivation of EGFR, which, in turn,
caused TEER changes (see SI Text, Figs. S6 and S7).
starved Caco-2 cells. The cells were lysed, immunoprecipitated using anti-EGFR
loading, the blots were stripped and reprobed for EGFR. Zonulin caused a
dose-dependent increase in EGFR phosphorylation that reached a plateau at 15
?g/mL. (B) Zonulin at 50 ?g/mL was incubated either alone (lane 2) or in the
presence of 5 ?M of the EGFR-selective PTK inhibitor AG1478 (lane 3) on serum-
(C) Zonulin, either alone or in the presence of 5 ?M AG1478, was applied to the
luminal side of C57BL/6 WT intestinal segments at a concentration of 50 ?g/mL,
and TEER was measured at baseline (open bars) and 90 min postincubation
presence of AG1478 (n ? 4 mice for each group). t0, time point t ? 0. (D) The
zonulin-induced EGFR phosphorylation was significantly reduced following
treatment with 2-chain mature HP2 (50 ?g/mL; lane 3) compared with single-
chain zonulin (lane 2). Lane 1 shows EGFR phosphorylation in cells treated with
(A) Zonulin at increasing concentrations was incubated on serum-
www.pnas.org?cgi?doi?10.1073?pnas.0906773106Tripathi et al.
In the current study, we have identified zonulin as the precursor of
HP2. Mature human HPs are heterodimeric plasma glycoproteins
composed of ?- and ?-polypeptide chains that are covalently
associated by disulfide bonds and in which only the ?-chain is
glycosylated (19). Unlike the ?-chain (36 kDa), the ?-chain exists
in 2 forms [i.e., ?1(?9 kDa) and ?2(?18 kDa)]. The presence of
one or both of the ?-chains results in the 3 phenotypes HP1–1,
HP2–1, and HP2–2. These HP variants evolved from a mannose-
?-chain containing a complement control protein and the ?-chain
a catalytically dead chymotrypsin-like serine protease domain
(21–24). Other members of the MASP family include a series of
plasminogen-related growth factors [e.g., EGF, hepatocyte growth
migration, and disruption of intercellular junctions. Despite this
multidomain structure, the only function assigned to HPs, to date,
is to bind Hb to form stable HP-Hb complexes, thereby preventing
Hb-induced oxidative tissue damage (25). No function has ever
proteins in that their precursor proteins, instead of being cleaved in
the trans-Golgi complex, are proteolytically processed by comple-
ment Cr1LP in the endoplasmic reticulum (13). Of interest, the
endoplasmic reticulum fraction was the cellular fraction in which
the highest zonulin concentrations were detected (9).
Because the key biological effect of zonulin is to regulate
intercellular TJ function (7, 9–11), we studied recombinant pre-
HP2 in IP assays. In a dose- and time-dependent manner, pre-HP2
reduced TEER across murine small intestinal mucosa both ex vivo
after cleavage into its 2 ?2- and ?-subunits further supports the
functions. Whether this functional divergence relates to conforma-
cleaved mature protein is under study. The importance of protein
conformation in dictating HP protein function is further supported
by the finding that zonulin cross-reactive anti-Zot Ab recognized
the HP1 ?1-chain under denaturing conditions (Figs. 1A and 2B)
but failed to recognize nondenatured HP1 (Fig. S2 A and B).
Combined, these data confirm the identity of zonulin as pre-HP2.
We previously reported that the NH2-terminal amino-acid
sequence of zonulin has striking similarities to the light chain of
human ?-globulins (7), a similarity also noted for HP (26).
Clearance of the HP-Hb complex can be mediated by the
monocyte/macrophage scavenger receptor CD163 (25). Clustal
W dendrogram analysis showed a region in the zonulin ?-chain
just upstream of the CD163 binding site with the following
?-globulin–like consensus motif: QLVE—V—P. Whether dis-
crepancies between our previously reported zonulin sequence
and the pre-HP2 sequence as related to this consensus motif are
attributable to intraspecies variability associated with a high
zonulin mutation rate or to our sequence error at that time
remains to be established.
Zonulin contains growth factor-like repeats. Like zonulin,
growth factors affect intercellular TJ integrity (27, 28). We now
show that the single-chain zonulin, but not its cleaved mature
form, transactivates EGFR via PAR2 and that its effect on
TEER is prevented by pharmacological inhibition of EGFR or
siRNA-induced PAR2silencing. This suggests that the growth
factor motif in the single-chain zonulin, but not in the mature
2-chain HP2, has the molecular conformation required to induce
TJ disassembly by indirect transactivation via PAR2. Whether
the EGF-like repeat in zonulin might also directly engage the
EGFR ectodomain remains to be established.
Gliadin, the environmental trigger of CD, reportedly repro-
duces the effects of EGF on the actin cytoskeleton (16). These
effects are very similar to the effects we reported for zonulin (7).
Gliadin binds to the CXCR3 chemokine receptor (29), and this
interaction is coupled to zonulin release from both intestinal
cells (9) and whole intestinal tissues (10). Hence, it is likely that
the gliadin-related EGF effects are mediated through zonulin
release. We also have identified intestinal bacterial colonization
both cause polarized luminal secretion of zonulin (8). Therefore,
we focused our studies on early zonulin action (i.e., its activity at
intestinal luminal side). This approach may appear counterin-
tuitive, given the observation that both EGFR and PAR2are
expressed basolaterally (3, 30). However, evidence exists that
they also are apically expressed (31). The fact that we have
demonstrated that zonulin exerts a permeating effect, both in ex
vivo and in vivo, when applied to the luminal aspect of the
acts basolaterally as well. When environmental triggers (i.e.,
bacteria, gluten) are present in the intestinal lumen, zonulin is
released from enterocytes, a process that is mediated, at least for
gliadin, by CXCR3 (29). Following zonulin release and the
subsequent increase in IP, these triggers can reach the submu-
cosa, where zonulin-expressing immune cells can secrete zonulin
to the basolateral side. A similar bilateral action has been
reported for mucosal mast cell protease II, another serine
protease that controls IP acting from both luminal and serosal
The role of both EGFR and PAR2 in regulating epithelial
permeability has been previously reported (33, 34). However,
our study provides previously undescribed evidence that the
2 receptors work cooperatively to regulate small intestine
We have previously reported that zonulin is up-regulated
during the acute phase of CD (9, 10). Using HP-specific primers,
we now report the previously undescribed expression of zonulin
mRNA in human intestine. Furthermore, real-time PCR exper-
iments showed that zonulin expression was increased in CD
patients compared with normal controls. The enhanced expres-
sion of zonulin correlated with disease activity, because CD
patients who were on a GFD showed mean values for zonulin
CD and normal controls. Interestingly, Papp et al. (35) recently
reported that a polymorphism in the HP gene represents a
previously undescribed genetic risk factor for CD development
and its clinical manifestations.
The human plasma levels of HPs are between 100 and 300 mg
mL. Almost 8% of HPs are secreted in their proform (36),
suggesting that under physiological circumstances, 80–208
?g/mL pre-HP2 is present in human plasma. Therefore, the
concentrations of zonulin used in this study are within physio-
logical range and are most likely indicative of the signaling
pathways activated when zonulin is up-regulated during patho-
logical processes. Besides CD, increased IP has been reported in
other autoimmune diseases, including T1D (11), systemic lupus
erythematosus (37), and ankylosing spondylitis (38), further
delineating the importance of the paracellular pathway in the
pathogenesis of autoimmune diseases. These findings, together
with the observation that zonulin is overexpressed during the
acute phase of several immune-mediated diseases and its block-
age prevents the onset of the autoimmune response, suggest that
zonulin contributes to the pathogenesis of these conditions,
opening previously undescribed paradigms in the pathobiology
and treatment options of immune-mediated diseases.
with CD were obtained from the Center for Celiac Research serum bank. All
samples were depleted of albumin and IgG using commercially available kits
(Enchant Life Science kit; Pall Corporation and IgG ImmunoPure immobilized
Tripathi et al.PNAS ?
September 29, 2009 ?
vol. 106 ?
no. 39 ?
analyzed by SDS/PAGE, 2-DE, and WB analysis.
Human HPs. HP1–1 and HP2–2 extracted from human plasma were purchased
from Sigma. HP SDS/PAGE, both monodimensional gel electrophoresis and
2-DE, WB, and MS analyses are described in detail in (SI Text). HP deglycosy-
lation was performed by addition of PNGase F according to the manufactur-
er’s instructions (Sigma).
Human Zonulin/Pre-HP2 Cloning and Expression in a Baculovirus Expression
System and Its Cleavage by Proteases. Recombinant zonulin/preHP2 protein
production using a baculovirus system and its purification are described in SI
the serine proteases indicated, resolved by SDS/PAGE, and then stained with
SimplyBlue SafeStain solution (Invitrogen). For generation of 2-chain HP2,
20 min at 25 °C. The beads were removed by centrifugation, and the effec-
tiveness of the removal of trypsin was confirmed by assay of trypsin peptidase
activity against the substrate Glu-Gly-Arg-pNA (Bachem BioScience).
Ex Vivo and In Vivo IP Studies. The effects of zonulin on ex vivo and in vivo IP
were determined as previously described (8, 14) and are reported in detail in
Zonulin Activation of EGFR. To determine whether zonulin can activate EGFR,
for increasing exposure times to serum-starved high EGFR-expressing Caco-2
cells. The cells were lysed and processed for WB analysis with anti-phospho
EGFR (Y1068) Ab (Cell Signaling Technology, Inc.) as previously reported (39).
inhibitor AG1478 (Calbiochem).
Knockdown of PAR2 Through RNA Interference. The methods used to silence
PAR2 are reported in detail in SI Text.
Zonulin Gene Sequencing and Quantification from Intestinal Tissue from Pa-
from the second or third portion of the duodenum from subjects undergoing
a diagnostic upper gastrointestinal endoscopy. Subjects included were 10
on treatment with a GFD for at least 6 months. All patients had clinical
indications for the procedure and gave their informed consent to undergo an
additional biopsy for the purpose of this study. The study protocol was
approved by the Ethics Committee of the University of Maryland. The small-
intestine biopsies were immediately collected in RNAlater RNA Stabilization
Reagent (Qiagen) and stored at ?20 °C until processed. Total RNA extraction,
cDNA synthesis, and real-time PCR are described in SI Text.
Statistical Analysis. All values are expressed as mean ? SE. The analysis of
differences was performed by 2-tailed Student’s t tests to test differences
was performed where appropriate. Values of P ? 0.05 were regarded as
ACKNOWLEDGMENTS. This manuscript was partially supported by National
Institutes of Health grant DK048373 (to A.F.).
1. Rook GA, Stanford JL (1998) Give us this day our daily germs. Immunol Today 19:113–
2. Arrieta MC, Bistritz L, Meddings JB (2006) Alterations in intestinal permeability. Gut
function in the pathogenesis of gastrointestinal autoimmune diseases. Nat Clin Pract
Gastroenterol Hepatol 2:416–422.
4. Wapenaar MC, et al. (2008) Associations with tight junction genes PARD3 and MAGI2
in Dutch patients point to a common barrier defect for coeliac disease and ulcerative
colitis. Gut 57:463–467.
5. Rescigno M, Lopatin U, Chieppa M (2008) Interactions among dendritic cells, macro-
phages, and epithelial cells in the gut: Implications for immune tolerance. Curr Opin
6. Fasano A (2000) Regulation of intercellular tight junctions by zonula occludens toxin
and its eukaryotic analogue zonulin. Ann N Y Acad Sci 915:214–222.
7. Wang W, Uzzau S, Goldblum SE, Fasano A (2000) Human zonulin, a potential modu-
lator of intestinal tight junctions. J Cell Sci 113(Pt 24):4435–4440.
8. El Asmar R, et al. (2002) Host-dependent zonulin secretion causes the impairment of
the small intestine barrier function after bacterial exposure. Gastroenterology
9. Drago S, et al. (2006) Gliadin, zonulin and gut permeability: Effects on celiac and
non-celiac intestinal mucosa and intestinal cell lines. Scand J Gastroenterol 41:408–
10. Fasano A, et al. (2000) Zonulin, a newly discovered modulator of intestinal permeabil-
ity, and its expression in coeliac disease. Lancet 355:1518–1519.
11. Sapone A, et al. (2006) Zonulin upregulation is associated with increased gut perme-
ability in subjects with type 1 diabetes and their relatives. Diabetes 55:1443–1449.
12. Bowman BH, Kurosky A (1982) Haptoglobin: The evolutionary product of duplication,
unequal crossing over, and point mutation. Adv Hum Genet 12:189–194.
13. Wicher KB, Fries E (2004) Prohaptoglobin is proteolytically cleaved in the endoplasmic
reticulum by the complement C1r-like protein. Proc Natl Acad Sci USA 101:14390–
14. Meddings JB, Swain MG (2000) Environmental stress-induced gastrointestinal perme-
ability is mediated by endogenous glucocorticoids in the rat. Gastroenterology
15. Sollid LM (2002) Coeliac disease: Dissecting a complex inflammatory disorder. Nat Rev
16. Barone MV, et al. (2007) Growth factor-like activity of gliadin, an alimentary protein:
Implications for coeliac disease. Gut 56:480–488.
17. Cenac N, et al. (2004) PAR2 activation alters colonic paracellular permeability in mice
via IFN-gamma-dependent and -independent pathways. J Physiol 558:913–925.
18. van der Merwe JQ, Hollenberg MD, MacNaughton WK (2008) EGF receptor transacti-
vation and MAP kinase mediate proteinase-activated receptor-2-induced chloride
a liver secretory glycoprotein. J Biol Chem 256:1055–1057.
20. Maeda N, Yang F, Barnett DR, Bowman BH, Smithies O (1984) Duplication within the
haptoglobin Hp2 gene. Nature 309:131–135.
21. Kurosky A, et al. (1980) Covalent structure of human haptoglobin: A serine protease
homolog. Proc Natl Acad Sci USA 77:3388–3392.
22. Nielsen MJ, et al. (2007) A unique loop extension in the serine protease domain of
haptoglobin is essential for CD163 recognition of the haptoglobin-hemoglobin com-
plex. J Biol Chem 282:1072–1079.
23. Polticelli F, Bocedi A, Minervini G, Ascenzi P (2008) Human haptoglobin structure and
function—A molecular modelling study. FEBS J 275:5648–5656.
24. Wicher KB, Fries E (2006) Haptoglobin, a hemoglobin-binding plasma protein, is
present in bony fish and mammals but not in frog and chicken. Proc Natl Acad Sci USA
and susceptibility to diabetic cardiovascular disease. Circ Res 92:1193–1200.
26. Hunt LT, Dayhoff MO (1972) The origin of the genetic material in the abnormally long
human hemoglobin and chains. Biochem Biophys Res Commun 47:699–704.
27. Hollande F, et al. (2001) HGF regulates tight junctions in new nontumorigenic gastric
epithelial cell line. Am J Physiol Gastrointest Liver Physiol 280:G910–G921.
integrity and function by hepatocyte growth factor. Invest Ophthalmol Visual Sci
29. Lammers KM, et al. (2008) Gliadin induces an increase in intestinal permeability and
zonulin release by binding to the chemokine receptor CXCR3. Gastroenterology
basolateral, but not the apical, surface of enterocytes in the human gastrointestinal
tract. Gut 39:262–266.
of the Gastrointestinal Tract, ed Johnson LR (Elsevier Academic, Oxford, UK), 4th Ed,
32. Jacob C, et al. (2005) Mast cell tryptase controls paracellular permeability of the
intestine. Role of protease-activated receptor 2 and beta-arrestins. J Biol Chem
33. Bueno L, Fioramonti J (2008) Protease-activated receptor 2 and gut permeability: A
review. Neurogastroenterol Motil 20:580–587.
34. Raimondi F, et al. (2008) Bile acids modulate tight junction structure and barrier
function of Caco-2 monolayers via EGFR activation. Am J Physiol Gastrointest Liver
35. Papp M, et al. (2008) Haptoglobin polymorphism: A novel genetic risk factor for celiac
disease development and its clinical manifestations. Clin Chem 54:697–704.
36. Misumi Y, Tanaka Y, Ikehara Y (1983) Biosynthesis, intracellular processing and secre-
tion of haptoglobin in cultured rat hepatocytes. Biochem Biophys Res Commun
37. Pavon EJ, et al. (2006) Proteomic analysis of plasma from patients with systemic lupus
alpha1 isoforms. Proteomics 6(Suppl 1):S282–S292.
38. Liu J, et al. (2007) Identification of disease-associated proteins by proteomic approach
in ankylosing spondylitis. Biochem Biophys Res Commun 357:531–536.
by EGF receptor kinase inhibitors. Science 242:933–935.
www.pnas.org?cgi?doi?10.1073?pnas.0906773106Tripathi et al.