LETTER
doi:10.1038/nature12729
Antibacterial membrane attack by a pore-forming
intestinal C-type lectin
SohiniMukherjee1,Hui Zheng2,MehabawG.Derebe1, KeithM.Callenberg3, Carrie L. Partch4,DarcyRollins1, Daniel C. Propheter1,
Josep Rizo5, Michael Grabe3{, Qiu-Xing Jiang2* & Lora V. Hooper1,6*
Human body-surface epithelia coexist in close association with
complex bacterial communities and are protected by a variety of
antibacterial proteins. C-type lectins of the RegIII family are bac-
tericidal proteins that limit direct contact between bacteria and the
intestinal epithelium and thus promote tolerance to the intestinal
microbiota1,2. RegIII lectins recognize their bacterial targets by
binding peptidoglycan carbohydrate1,3, but themechanismbywhich
they kill bacteria is unknown. Here we elucidate the mechanistic
basis for RegIII bactericidal activity. We show that human RegIIIa
(also known as HIP/PAP) bindsmembrane phospholipids and kills
bacteria by forming a hexameric membrane-permeabilizing oligo-
meric pore. We derive a three-dimensional model of the RegIIIa
pore by docking the RegIIIa crystal structure into a cryo-electron
microscopicmapof theporecomplex, andshowthat themodelaccords
with experimentally determined properties of the pore. Lipopoly-
saccharide inhibits RegIIIa pore-forming activity, explaining why
RegIIIa is bactericidal for Gram-positive but not Gram-negative
bacteria. Our findings identify C-type lectins as mediators of mem-
brane attack in the mucosal immune system, and provide detailed
insight into an antibacterial mechanism that promotes mutualism
with the resident microbiota.
RegIIIa damages the surfaces of Gram-positive bacteria1, indicating
that RegIIIamight target bacterial membranes. We assessed the capa-
city of RegIIIa to permeabilize bacterial membranes by quantifying
bacterial uptake of a membrane-impermeant fluorescent dye (SYTOX
green). RegIIIa increased SYTOX green uptake when added to the
Gram-positive species Listeria monocytogenes, indicating damaged
membranes (Fig. 1a, b). RegIIIa has an anionic amino-terminal pro-
segment that inhibits bactericidal activity (but not peptidoglycan bind-
ing) bydocking to theprotein core through charge–charge interactions4.
The pro-segment is removed by trypsin on secretion into the intestinal
lumen, yielding bactericidally active RegIIIa (ref. 4). Bactericidally
inactive pro-RegIIIa did not induce SYTOX green uptake, indicating
minimalmembrane permeabilization (Fig. 1a). Thus, RegIIIa permea-
bilizes the bacterialmembrane, and the pro-segment inhibits this activity.
To test directly for membrane disruption by RegIIIa we used lipo-
somes composedof 85%zwitterionicphospholipid (PC) and15%acidic
phospholipid (PS). The liposomes encapsulated carboxyfluorescein, a
fluorescent dye. RegIIIa induced rapid dye efflux from PC/PS lipo-
somes (Fig. 1c),whichwas reducedwhenPC-only liposomeswere used
(Fig. 1d, e). This indicates a preference for acidic phospholipids that is
consistent with the acidic lipid content of bacterial membranes5 and
with the salt sensitivity of RegIIIamembrane toxicity (Extended Data
Fig. 2a, b). These findings indicate that RegIIIa interactions with lipid
bilayers aremediated by electrostatic interactions. pro-RegIIIa yielded
a diminished rate of dye release (Fig. 1f), indicating that the pro-segment
inhibits membrane permeabilization.
Wenext assessedRegIIIa lipid-binding activity bymeasuring changes
in the intrinsic fluorescence of tryptophan residues6.Weobserved increased
tryptophan fluorescence intensity only when RegIIIa was added to
PS-containing liposomes (Fig. 1g–i), indicating that RegIIIa interacts
with acidic phospholipids. Furthermore, we observed fluorescence res-
onance energy transfer (FRET) between donor RegIIIa tryptophan
residues and dansyl-labelled PC/PS liposomes7 (Fig. 1j, k). FRET was
inhibited by the pro-RegIIIaN-terminal pro-segment (Fig. 1j, k), indi-
cating that the pro-segment inhibits bactericidal activity by hindering
lipid binding. Consistent with its inability to bind lipids, pro-RegIIIa
did not inhibit RegIIIa bactericidal activity in mixing experiments
(Extended Data Fig. 2c).
Several membrane-active toxins destabilize membranes by forming
monomeric or multimeric pores8. To test for RegIIIa pores, we per-
formed conductance studies in black lipidmembranes, amodel system
thatmimics the properties of a cellmembrane9. RegIIIa produced rapid
single-channel-like currents at 280mV in the presence of Mg21 ions
(Fig. 2a), with no current detected at 0mV. Using the Nernst–Planck
equationwe estimated the diameter of thepore at,12–14 A˚ (Extended
Data Fig. 3). The calculated pore size agreed with the lack of efflux
of fluorescein isothiocyanate-dextran-10 (FD10) or FD4, which have
Stokes diameters of ,44 A˚ and ,28 A˚, respectively (Fig. 2b). In con-
trast, carboxyfluorescein (,10 A˚) passed readily through the pores
(Figs 1c and 2b). These results show that RegIIIa forms functional
transmembrane pores and yield an estimate of the inner pore diameter.
When visualized by negative-stain electronmicroscopy (EM), numer-
ous circular structures of,100 A˚ diameter were observed on liposomes
incubated with RegIIIa (Fig. 2c and Extended Data Fig. 4a). Although
RegIIIa is a monomer in solution10, the size of the pores suggested that
theywereRegIIIamultimers.We therefore treated liposome-associated
RegIIIawith a crosslinking agent, solubilized the products in detergent,
and separated themby size-exclusion chromatography (Fig. 2d). In addi-
tion to a prominent monomer peak we detected a second, liposome-
dependent peak at a lower retention volume, indicating the formation
of a multimeric complex. Western blotting showed a single RegIIIa
species with mobility similar to that predicted for a hexamer (Fig. 2d),
suggesting that the pore was a RegIIIa hexamer.
After longer incubationswith lipid,RegIIIa formed filaments (Extended
Data Fig. 4b) similar to those in pancreatic secretions11. The filaments
were ,100 A˚ in diameter, correlating with the dimensions of the
RegIIIa pore (Fig. 2c). RegIIIa filamentation required lipid and was
dependent on RegIIIa pore formation, as pro-RegIIIa formed neither
pores nor filaments (Extended Data Fig. 4b, d). Filamentation partially
inhibited the ability of RegIIIa to permeabilize membranes (Extended
Data Figs 4c and 5a–c), as observed with other membrane toxic host
defence proteins where filamentation traps pore complexes and limits
damage to host cells12. These findings indicate that the RegIIIa filaments
*These authors contributed equally to this work.
1Department of Immunology, The University of Texas Southwestern Medical Center, Dallas, Texas 75390, USA. 2Department of Cell Biology, The University of Texas Southwestern Medical Center, Dallas,
Texas 75390, USA. 3Department of Biological Sciences, University of Pittsburgh, and Joint Carnegie Mellon University-University of Pittsburgh PhD Program in Computational Biology, Pittsburgh,
Pennsylvania 15261, USA. 4Department of Chemistry and Biochemistry, University of California, Santa Cruz, California 95064, USA. 5Department of Biochemistry and Department of Pharmacology, The
University of Texas Southwestern Medical Center, Dallas, Texas 75390, USA. 6The Howard Hughes Medical Institute, The University of Texas Southwestern Medical Center, Dallas, Texas 75390, USA.
{Present address: Cardiovascular Research Institute, University of California, San Francisco, San Francisco, California 94143, USA.
2 J A N U A R Y 2 0 1 4 | V O L 5 0 5 | N A T U R E | 1 0 3
Macmillan Publishers Limited. All rights reserved©2014
are higher-order assemblies of RegIIIa pore complexes and show that
filamentation limits RegIIIa toxicity.
Although the ,90-kDa RegIIIa pore complex was too small for
structure determination by single-particle cryoelectron microscopy
(cryoEM) methods13, the RegIIIa filaments were sufficiently large
for such analysis. We therefore reconstructed a three-dimensional
map of the RegIIIa filament and extracted the structure of theminimal
pore complex (Fig. 3a, b and Extended Data Fig. 6a–f). The nominal
resolution of our structure, 9.2 A˚, was limited by symmetry variability
and filament bending (Extended Data Fig. 6g–j and Supplementary
Information). Consistent with our crosslinking studies (Fig. 2d), the
minimal pore was a hexamer formed by three RegIIIa dimers related
by helical symmetry. The outer diameter of the pore assembly was
89 A˚, as observed by negative-stain EM (Fig. 2c). The pore height
was 55 A˚, sufficient to span a lipid bilayer (35–45 A˚)14. The inner
diameter was ,18 A˚, consistent with the pore size predicted by our
conductance measurements (Extended Data Fig. 3) and dye release
assays (Fig. 2b).
RegIIIa, like other epithelial bactericidal proteins such asa-defensins,
is constrained by disulphide bonds that prohibit large secondary struc-
ture changes onmoving froman aqueous to an apolar environment15,16.
This suggested the feasibility of docking the three-dimensional structure
of the RegIIIamonomer into the EM density map tomodel the organ-
ization of the pore complex further. First, we determined the crystal
structure of processed, bactericidally active RegIIIa (Extended Data
Fig. 7a) and compared it to the previously determined structure of
bactericidally inactive pro-RegIIIa. The two structures were similar,
although the amino acid side chains of the loop encompassing residues
93–99 (sequenceKSIGNSY) adopted different orientations in the active
RegIIIa structure (Fig. 3c). This was consistent with the conforma-
tional flexibility of this loop as indicated by a higher crystallographic
B-factor (Extended Data Fig. 7b).
The active RegIIIa structure could be docked into the cryo-EM
hexameric density map (Fig. 3d and Extended Data Fig. 6k, l), provid-
ing good spatial constraints for building a hexameric model. The
model indicates that the RegIIIa subunits in the pore assembly are
b c
0
50
100
0 2 4 6 8 10
RegIIIα (μM)
d
40
20
60
80
100
0
500 1,000
D
ye
r
el
ea
se
(%
o
f m
ax
)
Time (s)
Liposomes+:
No protein
RegIIIα
OG
Protein
0
e f
i
40
20
60
80
100
0
PC/PS PC PS
Lipid composition
***
***
NS
***
**
–RegIIIα
+RegIIIα
RegIIIα
Pro-RegIIIα
0
100
200
300
400
0 0.5 1.0
D
ye
r
el
ea
se
r
at
e
(fl
uo
re
sc
en
ce
u
ni
ts
s
–1
)
RegIIIα (μM)
***
g
0 100 200 300 400 500
0
5,000
10,000
15,000
20,000
Lipid conc. (μM)
Fl
uo
re
sc
en
ce
in
cr
ea
se
(A
U
)
Pro-RegIIIα
RegIIIα
a
RegIIIα
RegIIIα+PC/PS
RegIIIα+PC
320 340 360 380
0
50,000
100,000
150,000
200,000
250,000
300,000
Fl
uo
re
sc
en
ce
(A
U
)
Wavelength (nm)
j
h
S
Y
TO
X
u
p
ta
ke
(%
o
f m
ax
)
40
20
60
80
100
0
10 20 30 400
Time (min)
Untreated
RegIIIα
BSA
Pro-RegIIIα
0 500 1,000
Time (s)
40
20
60
80
100
0
OG
Protein
D
ye
r
el
ea
se
(%
o
f m
ax
) PC
PC/PS
PC+RegIIIα
PC/PS+RegIIIα
PS
PS+RegIIIα
Lipid composition:
0
50,000
100,000
150,000
200,000
250,000
320 340 360 380
0 μM
Lipid conc.:
50 μM
100 μM
500 μM
Wavelength (nm)
Fl
uo
re
sc
en
ce
(A
U
)
480 500 520 540 560
20,000
Fl
uo
re
sc
en
ce
(A
U
)
Wavelength (nm)
Dansylated
liposomes+:
Pro-RegIIIα
RegIIIα
No protein 30,000
40,000
50,000
60,000
k
0 1 2 3 4 5
Protein conc. (μM)
0.0
0.1
0.2
–0.1
FR
E
T
ef
fic
ie
nc
y
Pro-RegIIIα
RegIIIα0.3
S
Y
TO
X
u
p
ta
ke
(%
o
f m
ax
)
D
ye
r
el
ea
se
(%
o
f m
ax
)
Figure 1 | RegIIIa permeabilizes the bacterial membrane. a, Listeria
monocytogenes was treated with 25mM RegIIIa, pro-RegIIIa, or BSA or left
untreated, and bacterial uptake of SYTOX green was measured. Results are
representative of three independent experiments, and are expressed as a
percentage ofmaximumSYTOXuptake in thepresence of 0.2%SDS. b, SYTOX
green uptake by L. monocytogenes in the presence of increasing RegIIIa
concentrations. Assays were performed in triplicate. Means6 s.e.m. are
plotted. c, Carboxyfluorescein (CF)-loaded liposomes (10mM lipid; 85% PC/
15% PS) were treated with 1mMRegIIIa. 1.0% octylglucoside (OG) was added
towards the end to disrupt remaining liposomes. Dye efflux is expressed as
percentage of maximal release by detergent. Results are representative of five
independent experiments. d, 10mM RegIIIa was added to carboxyfluorescein-
loaded liposomes (100mM lipid; 100% PC, 100% PS or 85% PC:15% PS), and
dye efflux was monitored over time. Representative results are shown.
e, Averaged results from three independent replicates of the experiment
shown in d. NS, not significant; **P, 0.01; ***P, 0.001. f, Initial rate of
liposome dye efflux (100mM lipid) as a function of RegIIIa and pro-RegIIIa
concentration. Results are representative of three independent experiments.
*P, 0.05; **P, 0.01. g, Intrinsic tryptophan fluorescence of 1 mM RegIIIa
wasmeasured in the presence of increasing lipid concentrations. h, Tryptophan
fluorescence of 1 mM RegIIIa and pro-RegIIIa as a function of lipid
concentration. i, Intrinsic tryptophan fluorescence of 1mM RegIIIa was
measured in the presence of liposomes (100mM lipid) of varying lipid
composition. j, 5.0mMRegIIIa or pro-RegIIIawas added to liposomes (100mM
lipid) incorporating 5% dansyl-PE and dansyl fluorescence was monitored.
Assays were performed in triplicate. k, FRET efficiency as a function of RegIIIa
and pro-RegIIIa concentration. Assays were performed in triplicate.
Means6 s.e.m. are plotted.
RESEARCH LETTER
1 0 4 | N A T U R E | V O L 5 0 5 | 2 J A N U A R Y 2 0 1 4
Macmillan Publishers Limited. All rights reserved©2014
oriented with the carbohydrate-binding loop pointing towards the
central channel, and the loop encompassing residues 93–99 and the
N and carboxy (C) termini oriented towards the lipid bilayer (Fig. 3d).
The resolution of ourmap did not allow us to extract detailed informa-
tion about intermolecular interactions in the pore complex. There was
imperfect docking of the carbohydrate-binding loop, the loop encom-
passing residues 93–99, and the farN terminus (Fig. 3d), consistentwith
the conformational flexibility of these regions (Extended Data Fig. 7b).
We used mutagenesis to assess experimentally the orientation of
RegIIIa in the pore complex. Ourmodel predicts that the basic residue
Lys 93 is oriented towards the lipid bilayer (Fig. 3d) and thus might be
involved in interactions with the negatively charged phospholipids
required for RegIIIa–liposome interactions (Fig. 1d, e). A Lys93Ala
mutation, but not conservative Lys93Arg and Lys93His mutations,
reduced the toxicity of RegIIIa for liposomes as well as intact bacteria
(Fig. 3e, f and Extended Data Fig. 8a). In contrast, a Glu114Gln muta-
tion, which resides in the carbohydrate-binding loop (Fig. 3c)3, did not
have an impact on membrane toxicity, consistent with its predicted
position near the pore interior (Fig. 3d, e). As expected, the Lys93Ala
mutation but not the Glu114Gln mutation inhibited filament forma-
tion (ExtendedData Fig. 8b). Finally, the orientation of theN terminus
towards the lipid bilayer is consistent with the role of the N-terminal
pro-segment in inhibiting RegIIIa interactions with lipid and reducing
membrane toxicity (Fig. 1a, f, h, j, k).
We next calculated the energetics of pore insertion into a PC-likemem-
brane bilayer using physics-based computationalmodelling (Extended
Data Fig. 9a–d)17. The model predicts that basic residues are located
near the membrane–water interface whereas a strip of hydrophobic
and polar residues is buried in the membrane core (Fig. 3g). The
complex presents a positive electric field to the membrane (Extended
Data Fig. 9e, f), creating an unfavourable electrostatic energy unless
negatively charged PS-like lipids are added to the membrane (Fig. 3h).
This is consistent with our finding that PS lipids are necessary for
RegIIIa toxicity (Fig. 1d, e). Finally, calculations on the Lys93Ala
mutant showed reduced stability (Fig. 3h) due to loss of favourable
electrostatic interactions between Lys 93 and negatively charged lipids.
Thus, the model reveals that charge sequestration is a critical deter-
minant of RegIIIa pore stability in the membrane. Furthermore, the
model predicts that Arg 166 interacts with the membrane surface
(Extended Data Fig. 10a). Consistent with this prediction, an Arg166Ala
mutation reduced membrane toxicity of RegIIIa (Extended Data
Fig. 10b). In contrast, mutating Arg 39, which is exposed to aqueous
solvent in the model, had little effect on RegIIIa membrane toxicity
(Extended Data Fig. 10a, b). Thus, our model accurately predicts the
experimental behaviour of the RegIIIa pore.
RegIIIa selectively targets Gram-positive bacteria1, raising the ques-
tion of why RegIIIa cannot kill Gram-negative bacteria by permeabi-
lizing the outer membrane. In contrast to PC/PS liposomes, liposomes
composed of an Escherichia coli total lipid extract were not disrupted
by RegIIIa (Fig. 4a), indicating that a component of the lipid extract
inhibitedmembranepermeabilization. Lipopolysaccharide (LPS), amajor
constituent of the Gram-negative outer membrane, inhibited RegIIIa-
mediated liposome disruption and antibacterial activity (Fig. 4b, c),
indicating that LPS is one factor that prevents RegIIIa-mediated per-
meabilization of Gram-negative bacteria.
Finally, we postulated that the trypsin-cleavable inhibitory N ter-
minus of pro-RegIIIa evolved to suppress pore-forming activity and
thus minimize cytotoxicity during RegIIIa synthesis and storage in
epithelial cells. In support of this idea, RegIIIa was cytotoxic towards
cultured intestinal epithelial cells (MODE-K)18, and the pro-segment
suppressed this cytotoxicity (Fig. 4d, e).
Thus, RegIIIa kills its bacterial targets by oligomerizing on the bac-
terial membrane to form a membrane-penetrating pore (Extended
Data Fig. 1). Membrane attack by pore formation represents a previ-
ously unappreciated biological activity for theC-type lectin family.Our
findingsmay provide insight into the evolutionary origins of the lectin-
mediated complement pathway, in which recruited complement pro-
teins disrupt microbial membranes19. With its intrinsic capacity for
membrane attack, RegIIIamay represent a more evolutionarily prim-
itivemechanismof lectin-mediated innate immunity.We propose that
the lectin-mediated complement pathway could have evolved from a
directly bactericidal ancestral lectin, with the bacterial recognition
function retained by the descendent C-type lectin(s) and the mem-
brane attack function assumed by recruited accessory proteins that
assemble into the membrane attack complex.
c
d
0 mV 0 mV
–80 mV
50
p
A
5 s
a
b
FD4
Stokes diameter=~28 Å
Protein
0
20
40
60
80
100
D
ye
r
el
ea
se
(%
)
Protein
OG
FD10
Stokes diameter=~44 Å
0 2 4 6 8
0
50
100
150
200
250
Elution volume (ml)
m
A
U
Dimer
Tetramer
Hexamer
Predicted
mobilities: 100
75
50
25
37
20
15
kDa+liposomes –lipo.
10 12 14 16
67 43 13.7 6.5Size-exclusion
standards (kDa):
Monomer
0
100
200
300
400
–liposomes
+liposomes
m
A
U
Time (s)
Liposomes+:
No protein
5 μM RegIIIα
500 1,000
Protein
CF
Stokes diameter=~10 Å
0
0
20
40
60
80
100
0
20
40
60
80
100
100 Å
100 Å
100 Å
100 Å
Figure 2 | RegIIIa forms a transmembrane pore. a, RegIIIa-dependent
current flow across a planar lipid bilayer is depicted as a function of time.
No current was observed before the application of a voltage across the
membrane. Upon the application of 280mV, inward current was observed,
and returning the membrane potential to zero diminished the current because
the measured reverse potential was 24.0mV. The current trace is
representative of multiple independent experiments. b, Liposomes loaded with
FITC-Dextran 10 (FD10), FITC-Dextran 4 (FD4), or carboxyfluorescein (CF)
were treated with 5.0mM RegIIIa and dye release was monitored ove