Corrected: Publisher correction
ARTICLE
DOI: 10.1038/s41467-018-06086-4 OPEN
In vivo neutralization of dendrotoxin-mediated
neurotoxicity of black mamba venom by oligoclonal
human IgG antibodies
Andreas H. Laustsen 1, Aneesh Karatt-Vellatt 2, Edward W. Masters2, Ana Silvia Arias 3,
Urska Pus1, Cecilie Knudsen1, Saioa Oscoz3, Peter Slavny 2, Daniel T. Griffiths 2, Alice M. Luther 2,
Rachael A. Leah2, Majken Lindholm 2, Bruno Lomonte 3, José María Gutiérrez3 & John McCafferty 2
The black mamba (Dendroaspis polylepis) is one of the most feared snake species of the
African savanna. It has a potent, fast-acting neurotoxic venom comprised of dendrotoxins and
α-neurotoxins associated with high fatality in untreated victims. Current antivenoms are both
scarce on the African continent and present a number of drawbacks as they are derived from
the plasma of hyper-immunized large mammals. Here, we describe the development of an
experimental recombinant antivenom by a combined toxicovenomics and phage display
approach. The recombinant antivenom is based on a cocktail of fully human immunoglobulin
G (IgG) monoclonal antibodies capable of neutralizing dendrotoxin-mediated neurotoxicity of
black mamba whole venom in a rodent model. Our results show the potential use of fully
human monoclonal IgGs against animal toxins and the first use of oligoclonal human IgG
mixtures against experimental snakebite envenoming.
1 Department of Biotechnology and Biomedicine, Technical University of Denmark, Søltofts Plads 224, DK-2800 Kongens Lyngby, Denmark. 2 IONTAS Ltd.,
Iconix Park, London Road, Pampisford, Cambridgeshire CB22 3EG, United Kingdom. 3 Instituto Clodomiro Picado, Facultad de Microbiologiá, Universidad de
Costa Rica, San Jose ́ 11501-2060, Costa Rica. Correspondence and requests for materials should be addressed to A.H.L. (email: ahola@bio.dtu.dk)
or to A.K-V. (email: akv@iontas.co.uk) or to J.M.Gér. (email: jose.gutierrez@ucr.ac.cr)
NATURE COMMUNICATIONS |          (2018) 9:3928 | DOI: 10.1038/s41467-018-06086-4 | www.nature.com/naturecommunications 1
1234567890():,;
ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06086-4
Snakebite envenoming exacts a death toll of 80–150,000 results, thus, provide a proof of concept that oligoclonal mixturesvictims each year, leaves approximately four times as many of recombinant human IgG antibodies can be exploited to treatmaimed for life1, and has recently been recognized as a envenoming by the black mamba.
Neglected Tropical Disease by the World Health Organiza-
tion (WHO)2. Antivenoms derived from the plasma of hyper- Results
immunized animals remain the mainstay of snakebite enve- Description and preparation of venom antigens (toxins).
noming therapy3. However, these present a range of drawbacks D. polylepis venom was fractionated using RP-HPLC11, resolving
due to their relatively high cost and their heterologous nature that the key dendrotoxins in four venom fractions (Dp5, Dp6, Dp7,
causes, in some patients, a number of side effects, such as serum and Dp8) that cannot be further resolved in quantitative
sickness and early adverse reactions, which may include severe yields with standardized techniques. While Dp8 contains almost
anaphylaxis4–6. Furthermore, it is estimated that only a fraction pure dendrotoxin-1 (P00979 (https://www.uniprot.org/uniprot/
of the antibodies in most current antivenoms have a therapeutic P00979)), the venom fractions Dp5, Dp6, and Dp7 are ‘mixed
value, as the majority of antibodies isolated from animal plasma fractions’ that contain similar amounts of at least one den-
are directed against antigens that are either unrelated to snake drotoxin and at least one type II α-neurotoxins. Previous pro-
venom or related to venom components with negligible con- teomic studies have identified the toxin components of Dp5, Dp6,
tribution to venom toxicity5. The implication follows that and Dp7 to contain the same dendrotoxin (a homolog of den-
high amounts of antivenom protein may be needed to treat a drotoxin-δ, P00982 (https://www.uniprot.org/uniprot/P00982),
snakebite, with heterologous protein loads reaching as much as from the Eastern green mamba, D. angusticeps) and a type II α-
15 g/treatment for some antivenoms in severe envenoming cases7. neurotoxin (α-elapitoxin Dpp2c, P01397 (https://www.uniprot.
Particularly, elapid antivenoms often have an unbalanced anti- org/uniprot/P01397))11.
body content with relatively low amounts of antibodies against
small neurotoxic venom components that have low immuno-
genicity, which often leads to low immune responses in pro- Phage display selection and screening of scFv binders. Fol-
duction animals8–10. Despite the maturity of immunotherapy, lowing three rounds of panning against selected venom fractions
there remains a need for cost-effective antivenoms with improved containing dendrotoxins, polyclonal ELISAs revealed that anti-
safety and ef cacy5. body binders had been enriched (Fig. 1). Antibody genes (in scFvfi
The notorious black mamba (Dendroaspis polylepis) from format) were isolated from both the second and third panning19
sub-Saharan Africa is a particularly dangerous species due to its rounds, sub-cloned into an scFv bacterial expression vector , and20
size, defensive nature, and fast-acting neurotoxic venom. Life 188 clones were picked and their antibody expressed . Recom-
threatening clinical manifestations of D. polylepis envenoming binant monoclonal antibodies were tested for binding to respec-
include accid paralysis due to blockade of neuromuscular tive target antigens (see example with Dp8 as antigen Fig. 2).fl
transmission resulting from inhibition of nicotinic acetylcholine Using a cut-off score of 5000 fluorescence units (25 times above
receptors in the peripheral nervous system caused by α- the background binding signal), the top binders (up to 94) were
neurotoxins (both short-chain and long-chain types) of the picked for each antigen for further characterization (DNA
three-finger toxin superfamily11. In terms of abundance, however, sequencing and affinity ranking). An Expression-Normalized
the venom of D. polylepis is dominated by dendrotoxins, which Capture (ENC) assay was used to rank the antibody clones by
are a unique type of neurotoxins that inhibit voltage-dependent affinity. In this assay, limiting amounts of anti-FLAG antibodywere used to capture FLAG-tagged scFvs in the expression culture
potassium channels causing excitatory effects that result in
involuntary muscle contractions12. In the venom, short neuro- supernatant. Since the scFv expression level for each clone inculture supernatant is well above the capture capacity of the anti-
toxins, long neurotoxins, and dendrotoxins combine synergisti-
cally to provide D. polylepis with a potent neurotoxic bite11. FLAG antibody coated in each well, differences in antibodyexpression are normalized so that the binding signal better
Recently, recombinant antivenoms based on oligoclonal mix-
tures of human antibodies have been proposed as a cost- reflects differences in affinity between scFv and antigen.
competitive alternative to current antivenoms13. Additionally,
such recombinant antivenoms may provide safer and more Conversion of scFvs to IgG format and characterization. A
efficacious snakebite therapies due to their compatibility with panel of unique scFv-formatted antibodies that yielded the
the human immune system and the possibility of only including highest binding signals in the ENC assay were selected for con-
antibodies of therapeutic value, targeting medically relevant snake version to IgG format. Following expression by transient trans-
TM
venom toxins, in the antivenom mixture. To discover such anti- fection in Expi293 cells, ELISA was used to confirm retention
bodies, phage display has been identified as a promising tech- of target binding.
nology14 and has already yielded a number of neutralizing Twenty-five IgG-formatted antibodies were produced and
antibody fragments targeting venom toxins from snakes purified by protein A chromatography. Using a fluid-phase
(reviewed in Laustsen et al. 2016)5. However, to the best of our technique based on protein G-beads pull-down of antigen-
knowledge, no fully human IgG antibody has been reported antibody complexes, followed by acid dissociation and MALDI-
against any venom toxin from any multicellular organism, let TOF MS analysis (see example in Fig. 3), binding to a
11
alone a snake. Human IgGs have the benefits over antibody dendrotoxin homologous to dendrotoxin-δ in the venom
fragments of a prolonged half-life and different effector functions fractions was confirmed for 4/4 IgGs against Dp5, 3/4 IgGs
that depend on the Fc fragment. This may be of great therapeutic against Dp6, and 1/8 IgGs against Dp7 (Table 1). The Dp8
value for neutralization of systemically-acting toxins that leak fraction was known to consist predominantly of dendrotoxin-1,
15,16 with only minor traces of other toxins11from the bite site in victims over the course of days . Here, we . A pull-down experiment
report the discovery of a suite of human IgGs that provide pro- was performed for clone 367_01_H01, which confirmed that
tection in vivo against dendrotoxins from the black mamba when dendrotoxin-1 was indeed the target for this clone (Table 1).
administered by intracereberoventricular injection. This discovery
approach combined toxicovenomics17, antibody phage display In vivo neutralization of dendrotoxins. In total, 24 out of 25
technology18, antibody engineering, mammalian cell expression, recombinant human IgGs targeting black mamba neurotoxins
and whole venom in vivo neutralization studies in rodents. These were tested in vivo. All IgGs were evaluated for neutralization of
2 NATURE COMMUNICATIONS |          (2018) 9:3928 | DOI: 10.1038/s41467-018-06086-4 | www.nature.com/naturecommunications
NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06086-4 ARTICLE
Anti-Dp5 Anti-Dp7
8 8
6 Round 2 6 Round 2
Round 3 Round 3
4 4
2 2
0 0
p5 p6 p7 p8 bt
x ep ilk alr p5 p6 p7 p8 t
x ep ilk al
D D D D C tl-s
M b-
G D D D D Cb trl-s
M b-
G
Ga- -G
a
b b
Antigen Antigen
Anti-Dp6 Anti-Dp8
10 8
8 Round 2 6 Round 2
6 Round 3
4 Round 3
4
2 2
0 0
p5
l l
D Dp
6 p7 p8 bt
x p
re il
k a p5 p6 p7 p8 bt
x p ilk a
D D C t-s M
e
b-
G D D D D C tr-s Ml l b-
G
a
-G -G
a
b b
Antigen Antigen
Fig. 1 Polyclonal phage ELISA signals for the output phages for each selection. A significant increase in binding signal is observed from round 2 to 3 for all
selections, and for selections against Dp5, Dp6, and Dp7, a high degree of cross-reactive binding exists for the phages against all three antigens (Dp5, Dp6,
and Dp7). Negative control antigens: Cbtx α-cobratoxin, b-Gal β-galactosidase, Strep Streptavidin. All experiments were performed in triplicates on distinct
samples. Error bars represent standard deviations
a Polyclonal phage ELISA b Monoclonal scFv ELISA
4 20
18
16
3
14
12 Hits
2 10
8
6
1
4
2
0 0
p5 p6 p7 p8 txb Ga
l ep es scFv clones
D D D D rC tb- S ha
g
o 
p
N
Antigen
c DNA sequencing d Monoclonal IgG ELISA
12
8
90 hits sequenced
35 unique VH+VL CDR3 4
29 unique VH CDR3
0
Clone ID
Fig. 2 Schematic overview of selected results from the employed discovery process. Here, demonstrated for human IgGs against the antigen, Dp8.
a Polyclonal phage ELISA signals against different venom fractions and negative control antigens (Cbtx α-cobratoxin, b-Gal β-galactosidase, Strep
Streptavidin). b Monoclonal scFv ELISA signals against Dp8. c Summary of DNA sequencing results. Sequences are defined as unique based on VH and VL
CDR3 sequences. dMonoclonal IgG ELISA signals for converted clones demonstrating retained binding for the majority of the clones upon conversion from
the scFv format
NATURE COMMUNICATIONS |       (   2018) 9:3928 | DOI: 10.1038/s41467-018-06086-4 | www.nature.com/naturecommunications 3
Emission, 615 nm
× 100,000
Emission, 615 nm
Emission, 615 nm
× 100,000
× 100,000
Emission, 615 nmn Emission, 615 nm Emission, 615 nm Emission, 615 nm
× 10,000 × 10,000 × 100,000 × 100,000
366_01_A01
366_01_A07
366_01_B06
366_01_C01
366_01_C09
366_01_D04
367_01_D12
367_01_E03
367_01_F04
367_01_F06
367_01_F07
367_01_F08
367_01_F11
367_01_G05
367_01_G08
367_01_H01
367_01_H07
367_01_H09
ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06086-4
Table 1 Overview of all MALDI-TOF MS pull-down experiments
IgG Antigen Mass(es) detected (venom fraction) Mass(es) detected (whole venom) Corresponds to
(M+H) (M+H)
360_01_B12 Dp5 6555.1 *Ca. 6500+ 10465.9 “Dendrotoxin-δ“
360_01_C09 Dp5 6555.6 6554.8 “Dendrotoxin-δ“
361_01_F07 Dp5 6556.5 6554.6 “Dendrotoxin-δ”
361_01_G08 Dp5 6556.2 6555.8 “Dendrotoxin-δ“
362_01_A08 Dp6 6556.7 6554.5 “Dendrotoxin-δ“
362_01_D01 Dp6 6557.4 6554.2 “Dendrotoxin-δ“
363_01_F07 Dp6 6557.6 6554.3 “Dendrotoxin-δ“
363_01_G12 Dp6 8069.7 *Ca. 8000+ ca. 10400 Type II α-neurotoxin
364_01_A04 Dp7 8002.6+ 8067.0 8007.2+ 8068.2 Type II α-neurotoxin
364_01_A01 Dp7 8005.2+ 8024.8+ 8069.7 8008.2+ 8066.4 (+*ca. 10400) Type II α-neurotoxin
364_01_B01 Dp7 8002.2+ 8065.9 8005.9+ 8067.3 Type II α-neurotoxin
364_01_C11 Dp7 8006.1+ 8066.9 N.t. Type II α-neurotoxin
364_01_D03 Dp7 8005.5+ 8025.4 *(−) Type II α-neurotoxin
364_01_D04 Dp7 8004.8+ 8069.3 *Ca. 8000 Type II α-neurotoxin
365_01_F03 Dp7 8005.4+ 8028.8+ 8068.5 *(−) Type II α-neurotoxin
365_01_G06 Dp7 6560.3 6553.8 “Dendrotoxin-δ“
367_01_H01 Dp8 N.t. 7127.4 Dendrotoxin-1
*indicates a poor signal-to-noise ratio. N.t. not tested. (−)= could not be determined. M+H= determined molecular mass+ proton. “Dendrotoxin-δ“ denotes a D. polylepis dendrotoxin homologous to
the D. angusticeps dendrotoxin-δ. Instrumental error is within 0.02% of the observed mass values
a b
100 6563.1 2.5E+ 100 7128.9Dp5 8450.090 90 Venom
80 Dtx 80 6555.0
70 8015.5 70 3FTx
60 6523.8 60 Dtx
50 50 7165.9
40 3FTx 40
30 30 6467.0
20 20 3557.04006.9 6760.0
10 10 8009.2
0 0
3000 4800 6600 8400 10,200 12,000 3000 4800 6600 8400 10,200 12,000
100 6556.4 7568.0 100 6554.7 5454.0
90
361_01_F07 90 361_01_F07
80 Pull-down 80 Pull-down
70 70
60 Dtx 60
50 50 Dtx
40 40
30 30 3130.1
20 20 3881.7
10 10 10465.7
0 0
3000 4800 6600 8400 10,200 12,000 3000 4800 6600 8400 10,200 12,000
c d
100 6555.6
7128.9
3.6E+ 100 8450.0
90 Dp6 90 Venom
80 Dtx 80
6555.0
70 70 Dtx 3FTx
60 60
50 4030.8 3FTx 50 7165.9
40 8069.9 40
30 3273.0 30
6467.0
6613.9
20 3557.04400.4 20 6760.0
10 3304.7 7010.0 10
8009.2
0 0
3000 4800 6600 8400 10,200 12,000 3000 4800 6600 8400 10,200 12,000
100 6557.6 8858.0 100 6554.3 3262.0
90
363_01_F07 90 363_01_F07
80 Pull-down 80 Pull-down
70 70
60 Dtx 60
50 50 Dtx
40 40
30 30
20 20 3127.8 3881.2
10 10 10464.4
0 0
3000 4800 6600 8400 10,200 12,000 3000 4800 6600 8400 10,200 12,000
Fig. 3 Example of MALDI-TOF MS analysis of antigen-antibody complex pull-down experiments. a IgG 361_01_F07 pull-down from venom fraction Dp5.
b IgG 361_01_F07 pull-down from whole venom. c IgG 363_01_F07 pull-down from venom fraction Dp6. d IgG 363_01_F07 pull-down from whole venom.
Dtx Dendrotoxin, 3FTx Three-finger toxin
lethality by the intracerebroventricular (i.c.v.) route, where nine although most of these IgGs showed prolonged survival time, as
showed full (100%) protection against the venom fraction they compared to controls, during the assay. Eight IgGs provided
were raised against (Tables 2 and 3). Even at the highest dose partial survival in the 24 h assay at one or more dose regimes
tested, seven IgGs failed to provide survival in the 24 h assay, (Tables 2 and 3).
4 NATURE COMMUNICATIONS |          (2018) 9:3928 | DOI: 10.1038/s41467-018-06086-4 | www.nature.com/naturecommunications
NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06086-4 ARTICLE
Table 2 In vivo neutralization results for monoclonal IgG antibodies raised against Dp5, Dp6, and Dp7
Venom fractions
IgG ID Dp5 Dp6 Dp7
0.5:1 0.75:1 1:1 2:1 3:1 4:1 0.5:1 0.75:1 1:1 2:1 3:1 4:1 6:1 2:1 3:1 4:1 6:1
360_01_B12 0/4 4/4 4/4 1/4 1/4
360_01_C09 0/4 1/4 4/4 0/4 1/4
361_01_F07 2/4 3/4 3/4 4/4 4/4 3/4 2/4 1/4
361_01_G08 0/4 1/4 2/4 2/4 4/4 4/4 1/4 0/4
362_01_A08 3/4 3/4 4/4 4/4 4/4 1/4 2/4 3/4 3/4 0/4
362_01_D01 2/4 3/4 3/4 3/4 3/4 0/4 2/4 2/4 4/4 1/4
363_01_F07 4/4 4/4 4/4 4/4 4/4 0/4 0/4 2/4 4/4 4/4 4/4 1/4
363_01_G12 0/4 0/4
364_01_A01 0/4 0/4* 0/4 0/4
364_01_A04 0/4 0/4 1/4 3/4
364_01_B01 1/4* 1/4 1/4
364_01_C11 1/4
364_01_D03 0/4* 0/4 0/4
364_01_D04 0/4* 1/4 0/4
365_01_G06 4/4 3/4 4/4 4/4 4/4 0/4 0/4 3/4 0/4 3/4 4/4
Monoclonal IgGs were tested against individual venom fractions using i.c.v administration. Numbers indicate survival ratios at 24 h. * denotes that intravenous (i.v.) administration was used instead of i.c.
v. administration. Ratios are provided as molar ratios between IgG:toxin.
Table 3 In vivo neutralization results for monoclonal IgG antibodies raised against Dp8 and oligoclonal IgG cocktails
Venom fractions
IgG ID Dp8 Whole venom
2:1 4:1 6:1 1:1 2:1 3:1 4:1
366_01_A01 0/3 0/3 1/3
366_01_B06 0/3 0/3 0/3
366_01_C01 0/3 0/3 0/3
366_01_C09 2/3 2/3 0/3
367_01_F04 0/3 0/3 0/3
367_01_F07 1/3 0/3 3/3
367_01_F11 2/3 0/3 2/3
367_01_H01 3/3 3/3 3/3 1/4
367_01_H09 0/3 1/3 1/3
Cocktail 1 0/4 1/4 4/4
Cocktail 2 1/4 4/4
Cocktail 3 2/4
Cocktail 4 2/4
Monoclonal IgGs and IgG cocktails were tested against individual venom fractions and whole venom using i.c.v administration. Numbers indicate survival ratios at 24 h. Ratios are provided as molar ratios
between IgG:toxin.
The i.c.v. assay is particularly useful for assaying toxicity of all challenged mice died between 12–18 h. The present findings
dendrotoxins, as these neurotoxins are highly potent when thus demonstrate that effective dendrotoxin-targeting human
administered i.c.v., but display lower toxicity by intravenous (i.v.) IgGs were discovered, but more work (such as affinity matura-
administration, requiring relatively high doses to induce leth- tion) may be needed to improve the discovered IgGs that target
ality11. In contrast, α-neurotoxins are less potent when adminis- type II α-neurotoxins. Moreover, this work also demonstrates that
tered i.c.v. and are better assayed using the i.v. route of the discovered human IgGs could neutralize the lethal effect of
administration. Three of the four investigated venom fractions the target dendrotoxins present in more than one venom fraction
contain a mixture of dendrotoxins and type II α-neurotoxins11, (e.g., 363_01_F07 which completely neutralizes the dendrotoxins
and it was observed that only dendrotoxin-targeting IgGs were in Dp5 and Dp6 and provides some protection against Dp7)
able to provide full survival in the i.c.v. assay (compare (Table 2). The mass spectrometry data (Table 1) suggests that, in
Table 1 with Tables 2 and 3). Clones 364_01_A01, 364_01_B01, this case, the observed neutralization ability is due to the presence
364_01_D03, and 364_01_D04 target the type II α-neurotoxin of a key dendrotoxin (or very similar dendrotoxins) in more than
present in Dp7 and were therefore also assayed using the i.v. one venom fraction.
route. Unfortunately, these IgGs failed in providing full protec- To explore whether the dendrotoxin-mediated neurotoxicity of
tion, although clone 364_01_B01 provided a low survival rate (1/ the whole venom could be completely abrogated using the
4) when mice were challenged with 10.6 µg toxin pre-incubated discovered human IgGs, antibody cocktails were designed
with the IgG at a molar ratio of 3:1 (IgG:toxin). Similarly, clone (Table 4) and evaluated against whole venom by the i.c.v. route.
364_01_A01 provided significantly prolonged survival, although In all cases, the combination of antibodies to dendrotoxin-1 and
NATURE COMMUNICATIONS |        (  2018) 9:3928 | DOI: 10.1038/s41467-018-06086-4 | www.nature.com/naturecommunications 5
ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06086-4
the dendrotoxin-δ homolog was superior to anti-dendrotoxin-1 Finally, Cocktail 1 and 2 were tested against whole venom by
(367_01_H01) alone. Anti-dendrotoxin-1 (367_01_H01) was the i.v. route to demonstrate that these cocktails were truly
used along with different combinations of antibodies to dendrotoxin-specific and that whole venom cannot be neutralized,
dendrotoxin-δ. For example, Cocktail 2 included 363_01_F07 when using the i.v. route, unless both dendrotoxins and α-
and 365_01-G06 whereas Cocktail 1 included these, as well as neurotoxins in the venom are neutralized21. As expected,
an additional antibody (361_01_F07) (Tables 3 and 4). Both similar survival curves (rapid death upon injection) were observed
cocktails successfully provided full protection against whole for mice challenged with 25.8 µg of whole venom alone and mice
venom when injected via the i.c.v. route at a challenge dose of challenged with a similar dose of whole venom pre-incubated
1.5 µg of whole venom pre-incubated with the IgG cocktails at with Cocktail 1 and 2 at molar ratios of 3:1 and 4:1 (IgG:toxin),
IgG:toxin molar ratios of 4:1 and 3:1 (Fig. 4). Omission of one respectively.
or other of the anti-dendrotoxin-δ-homolog antibodies from For comparative purposes, an equine-derived F(ab’)2 poly-
Cocktail 2 yielded reduced protection. For example, neutraliza- clonal antivenom was tested for its efficacy against the lethal
tion tests against whole venom were performed i.c.v. using effect of D. polylepis venom by the i.c.v. route. This antivenom
Cocktail 3 (where 365_01_G06 had been substituted with had been previously shown to be highly effective in the
additional 363_01_F07), Cocktail 4 (where 363_01_F07 had neutralization of lethality of this venom by the i.v. route, with a
been substituted with additional 365_01_G06) (Table 4), and Median Effective Dose of 5.25 mg venom neutralized per mL
367_01_H01 alone. As seen in Fig. 5, none of these cocktails antivenom11, an observation that was confirmed in the present
were able to provide equivalent protection at equimolar doses study. In contrast, when lethality was tested by the i.c.v. route,
compared to Cocktail 2. Thus, Cocktail 2 represented the the antivenom failed to neutralize this venom even at a ratio
minimum cocktail providing full protection in this experiment. of 0.33 mg venom per mL antivenom, as all mice receiving the
mixture of venom and antivenom died, whereas control mice
injected with antivenom alone survived.
Table 4 Composition of the IgG cocktails Discussion
The results presented here are the first report of the use of human
anti-Dp5 Anti-Dp6 Anti-Dp7 Anti-Dp8 IgG antibodies capable of neutralizing animal toxins in vivo.
361_01_F07 363_01_F07 365_01_G06 367_01_H01 Moreover, with this report, we demonstrate that the dendrotoxin-
Cocktail 1 1 1 1 6 mediated neurotoxicity of whole venom of the black mamba can
Cocktail 2 2 1 6 be neutralized in an i.c.v. rodent model using carefully selected
Cocktail 3 3 6
Cocktail 4 3 6 oligoclonal mixtures of monoclonal human IgGs. Our results
further indicate that individual monoclonal dendrotoxin-
Numbers in table represent relative molar ratios between each IgG component targeting IgGs cannot achieve this alone, and it is likely to be
essential to employ antibody mixtures that neutralize several key
toxins of black mamba whole venom to achieve full protection
a Cocktail 1 (4 IgGs) b Cocktail 2 (3 IgGs)
100 100
80 Control 80
60 4:1 60 Control
40 40 3:12:1
20 1:1 20 2:1
0 0
0 3 6 9 12 15 18 21 24 0 3 6 9 12 15 18 21 24
Time (h) Time (h)
Fig. 4 Kaplan-Meier survival curves for antibody cocktails. Here, shown for a Cocktail 1 and b Cocktail 2 at different molar ratios against black mamba
whole venom administered i.c.v., demonstrating full protection at an IgG:toxin molar ratio of 4:1 for Cocktail 1 and 3:1 for Cocktail 2. Cocktail 1 contains the
IgGs: 361_01_F07 (anti-Dp5), 363_01_F07 (anti-Dp6), 365_01_G06 (anti-Dp7), 367_01_H01 (anti-Dp8). Cocktail 2 contains the IgGs: 363_01_F07 (anti-
Dp6), 365_01_G06 (anti-Dp7), 367_01_H01 (anti-Dp8). Each individual curve represents survival of a cohort of 4 animals
a Cocktail 3 b Cocktail 4 c 367_01_H01
100 100 100
90 90 90
80 80 80
70 70 70
60 60 60
50 Control 50 Control 50 Control
40 40 40
30 363_01_F07+ 30 365_01_G06+ 30 367_01_H01
20 367_01_H01 20 367_01_H01 20
10 10 10
0 0 0
0 3 6 9 12 15 18 21 24 0 3 6 9 12 15 18 21 24 0 3 6 9 12 15 18 21 24
Time (h) Time (h) Time (h)
Fig. 5 Kaplan–Meier survival curves for antibody cocktails. Here, shown for a Cocktail 3, b Cocktail 4, and c clone 367_01_H01 against black mamba whole
venom administered i.c.v. at an IgG:toxin molar ratio of 3:1. Cocktail 3 contains the IgGs: 363_01_F07 (anti-Dp6) and 367_01_H01 (anti-Dp8). Cocktail 4
contains the IgGs: 365_01_G06 (anti-Dp7) and 367_01_H01 (anti-Dp8). Each individual curve represents survival of a cohort of 4 animals
6 NATURE COMMUNICATIONS |          (2018) 9:3928 | DOI: 10.1038/s41467-018-06086-4 | www.nature.com/naturecommunications
Survival %
Survival %
Survival %
Survival %
Survival %
NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06086-4 ARTICLE
NheI NotI NcoI XhoI
EF1-alpha CMV 
promoter promoter
Leader VL CL Leader VH CH1 CH2 CH3
pINT3-hg1 vector
Fig. 6 Schematic representation of the genetic elements present in the pINT3-hg1 vector
against dendrotoxin-mediated neurotoxicity. Finally, through mL/min using Solution A (water, containing 0.1% TFA) and a gradient towards
systematic testing, we determined a minimum IgG cocktail solution B (acetonitrile, containing 0.1% TFA): 0% B for 5 min, 0–15% B over 10
(Cocktail 2) containing three human IgGs that is capable of min, 15–45% B over 60 min, 45–70% B over 10 min, and 70% B over 9 min.
Fractions were collected manually and dried in a vacuum centrifuge11.
providing full protection against lethality in mice co-injected with
lethal doses of whole venom by the i.c.v. route. The value of these
Biotinylation and MS analysis of toxins. Venom fractions were dissolved in
antibodies is underscored by the fact that a polyvalent antivenom phosphate buffered saline (PBS, Dulbecco’s Phosphate Buffered Saline; Sigma-
in clinical use, which is highly effective in the neutralization of D. Aldrich) to yield concentrations of 0.85–6.39 μg/μL. Biotin linked to N-hydro-
polylepis venom when tested by the i.v. route, failed to neutralize xysuccinimide (NHS) via a PEG4-linker (EZ-Link™ NHS-PEG4-Biotin, No-Weigh™
venom toxicity when assays were performed by the i.c.v. route. Format, Thermo Scientific, 21329) was added to the toxin solutions at molar ratios
Since dendrotoxins are the predominant neurotoxins upon i.c.v. of 1:1 to 1:2 (toxin: biotinylation reagent) and left at room temperature for 30 min.Buffer exchange columns (Vivacon 500, Sartorius, 2000 Da Molecular Weight Cut-
injection, whereas α-neurotoxins predominantly mediate toxicity Off) were employed for purification of the biotinylated toxins using three washes of
when using the i.v. route, our findings suggest that the equine 500 μL PBS and an elution volume of 150 μL PBS. Protein concentrations were
polyvalent antivenom has limitations for the neutralization of determined based on individually calculated extinction coefficients (http://web.
dendrotoxins, hence reinforcing the potential of recombinant expasy.org/protparam/) and absorbances measured on a BMG labtech PHERAStarFluorescence Spectrophotometer. The extent of biotinylation was analyzed by
antibodies against this type of neurotoxin. MALDI-TOF in a Proteomics Analyzer 4800 Plus mass spectrometer (Applied
Three limitations exist for this work. (1) When using an i.c.v. Biosystems) to ensure that over-biotinylation had not taken place23.
model, dendrotoxins are the predominant toxic components, and
neutralization of whole D. polylepis venom can be achieved by Phage display selection and primary screening. For phage display selection, the
neutralizing these toxins only. When switching to an i.v. model, IONTAS phage display library was employed. This library is a human antibody
10
neutralization of α-neurotoxins is likely to be essential, as these phage display library of 4 × 10 clones, with antibodies in the form of single chain
variable fragments (scFvs), which was constructed from B lymphocytes collected
are expected to be the main drivers of toxicity by this route. This from 43 non-immunized human donors24. Selection of scFv binders from the
speculation is supported by our finding that both IgG cocktails IONTAS phage display library and primary TRF assay screening for dendrotoxin
tested in this work could not provide survival against whole binders were performed as described elsewhere24,25. Briefly, selected antibodies
venom when i.v. administration was employed. (2) A model were sub-cloned from the phage display vector using Nco I and Not I restriction
endonuclease sites into a vector for expression of soluble scFvs19 and transformed
based on i.c.v. does not reflect a typical envenoming, but only into E. coli strain BL21 (DE3) (New England Biolabs). Individual scFv clones (up to
demonstrates the ability of toxin-targeting IgGs to neutralize 188 scFv clones against each target) were picked, expressed in 96-well format, and
some neurotoxins. Thus, despite the usefulness of the i.c.v. route scFv-containing supernatants tested for binding to their corresponding venom
to assess the neutralization of dendrotoxins, the use of the i.v. fraction targets (1–5 μg/mL) indirectly immobilized on streptavidin (10 μg/mL)
route is recommended to test neutralization of α-neurotoxins and coated MaxiSorp plates using the DELFIA system
25. For binding detection, 1 in
1500 dilution of anti-FLAG M2 (Sigma, F1804) conjugated with Europium (custom
crude venoms. (3) Finally, although incubation and co- labeled by Perkin Elmers) was used. Ninety-four binders against each target were
administration of venom and IgGs follows the recommended cherry-picked and sequenced (Eurofin Genomics sequencing service) using S10b
WHO guidelines22, such experiments fail to account for the primer (GGCTTTGTTAGCAGCCGGATCTCA). The antibody framework and
impact of toxicokinetics and pharmacokinetics. It would be CDR regions were annotated and analyzed to identify unique clones.
relevant to follow up on these observations with ‘rescue’ experi-
ments, where IgGs are administered i.v. after a period of time Expression-normalized capture (ENC) assay. For the ENC assays, black Max-
iSorp plates (Nunc) were coated overnight with anti-FLAG M2 antibody (Sigma,
following administration of whole venom i.m. or s.c., as this will 2.5 μg/ml in PBS, 50 μL per well). After blocking with 2% M-PBS (skim milk in
better reflect a real-life envenoming situation. PBS), washing with PBS, and addition of 30 μL of 6% M-PBS to each well, 30 μL of
Taken together, the data provided here constitutes a proof of individual auto-induction culture supernatants20 containing expressed scFv was
concept for the use of oligoclonal mixtures of recombinant added for each scFv to the assay plate. Plates were washed three times with PBS-T
(PBS, 0.1% Tween-20) and three times with PBS. Binding of biotinylated antigen
human IgGs against snakebite envenoming. The use of carefully (tested using both 2.5 nM and 25 nM of each antigen in 2% M-PBS, 50 μL per well
selected human IgG antibodies holds the promise of delivering for 1 h) was detected using Europium-labeled Streptavidin (Perkin Elmer, 1244-
safer and more effective treatments against snakebite envenoming 360, 1 μg/mL in PBS-M, 50 μL per well for 30 min).
due to the compatibility with the human immune system and the
possibility of only including antibodies of therapeutic value in a IgG expression and purification. VH and VL genes of 30 antibodies that showed
recombinant antivenom5,6. Moreover, as cost-efficacy remains highest binding signal in the ENC assay were sub-cloned into a dual promoter
one of the main challenges of delivering biotherapeutics to mammalian expression vector. The pINT3-hg1 vector (Fig. 6) has a dual promoter
1 expression cassette, in which the heavy chain expression is controlled by thepatients in developing countries , the use of oligoclonal recom- cytomegalovirus (CMV) promoter, and the light chain expression is driven by
binantly expressed human IgGs has been theorized to be cost- elongation factor-1 alpha (EF1-alpha) promoter. An alternative vector (pINT54-
competitive with current plasma-derived antivenoms13. These hg1) with a dual CMV promoter driving both heavy and light chain expression was
results bring hope that recombinant antivenoms based on fully also available and used in some cases. VH chains of the selected antibodies were
amplified from the pSANG10 scFv vector using primers pSang10_pelB
human IgG antibodies may in the future be available for the (CGCTGCCCAGCCGGCCATGG) and HLINK3_R (CTGAACCGCCTCCACCA
therapy against snakebite envenomings. CTCGA), and the VL chains were amplified using primers LLINK2_F
(CTCTGGCGGTGGCGCTAGC) and 2097_R (GATGGTGATGATGATGTGCG
GATGCG). Amplified VH genes were digested with Nco I and Xho I restriction
Methods endonucleases, and the VL genes were digested with Nhe I and Not I restriction
Venom fractionation. Pooled D. polylepis venom from several specimens origi- endonucleases. Digested VH and VL gene fragments were ligated into the pINT3-
nating from Kenya was obtained in lyophilized form from Latoxan SAS, France. Hg1 or pINT54 hg1 plasmids (digested with Nhe I and Xho I restriction endo-
Venom fractions Dp5, Dp6, Dp7, and Dp8 containing dendrotoxins from nucleases) along with a stuffer fragment (digested with Nco I and Not I restriction
D. polylepis were isolated from crude venom by RP-HPLC (Agilent 1200) on a C18 endonucleases) encoding the constant light chain region (CL) and the CMV pro-
column (250 × 4.6 mm, 5 μm particle; Teknokroma). Elution was carried out at 1 moter as a four-part ligation using T4 DNA ligase (Roche, 10481220001).
NATURE COMMUNICATIONS |          (2018) 9:3928 | DOI: 10.1038/s41467-018-06086-4 | www.nature.com/naturecommunications 7
ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06086-4
In order to express the IgG antibodies in mammalian cells, transfection quality injected with venom fractions/whole venom in PBS. Time of death was recorded,
DNA was prepared using Plasmid Plus Kit (Qiagen, 12945). A total of 180 μg of the and Kaplan–Meier curves were used to represent mouse survival.
DNA was transfected into 180 mL of Expi293TM cells (Thermo Fisher) using
ExpiFectamineTM 293 Transfection Kit (Thermo Fisher, A14525) according to
manufacturer’s instructions. Cells were harvested after 6 days of incubation at 37 ° Data availability
C, 5% CO2, 130 rpm on a 25 mm orbital shaker. Antibodies were purified from the The data that supports the findings of this study are available from the corresponding
culture supernatants by protein A chromatography using the Äkta Pure system authors upon reasonable request. Figures 1, 2, 3, 4, and 5 have associated raw data.
(GE Healthcare). The initial purification was performed using HiTrap MabSelect
SuRe 5 mL column (GE Healthcare, 11-0034-94), using 0.1 M Citrate buffer (pH Received: 24 May 2018 Accepted: 17 August 2018
3.0) for elution. Eluted proteins were neutralized with half the volume of 1M Tris
(pH 8.0) and dialyzed twice against 4 L of 2× PBS at 4 °C using GeBAflex Maxi Published online: 02 October 2018
dialysis tubes (Generon, D035). Dialyzed proteins were concentrated using Amicon
ultra centrifugation filters (Merck Millipore, UFC905024) according to
manufacturer’s instructions. The functionality of purified IgGs was confirmed by a
TRF binding assay. In this assay, IgG binding to biotinylated venom fractions References
immobilized on streptavidin coated Nunc MaxiSorp plates was detected using an 1. Gutiérrez, J. M. et al. Snakebite envenoming. Nat. Rev. Dis. Prim. 3, 17063
anti-human antibody conjugated with Europium (Perkin Elmers, 1244-330, 1 in (2017).
1000 dilution). 2. Chippaux, J.-P. Snakebite envenomation turns again into a neglected tropical
disease! J. Venom. Anim. Toxins Trop. Dis. 23, 38 (2017).
Identification of target toxin for IgGs. Each monoclonal IgG was mixed, at a 4:1 3. Gutiérrez, J. M., León, G., Lomonte, B. & Angulo, Y. Antivenoms for snakebite
molar ratio (IgG:toxin), with 0.5 µg of the venom fraction which it was envenomings. Inflamm. Allergy Drug. Targets 10, 369–380 (2011).
selected against, or with whole venom, and added to a 20 µL slurry of protein G- 4. Gutiérrez, J. M., Williams, D., Fan, H. W. & Warrell, D. A. Snakebite
agarose beads (Sigma P7700), in 50 µL of 0.05 M Tris, 0.75 M KCl, pH 7.0 envenoming from a global perspective: Towards an integrated approach.
buffer containing 2% bovine serum albumin. The mixtures were incubated for Toxicon 56, 1223–1235 (2010).
30 min at room temperature in a Thermomixer (Eppendorf) at 700 rpm. After 5. Laustsen, A. H. et al. From fangs to pharmacology: the future of snakebite
centrifugation for 15 s at 5000×g, the beads were washed twice with 200 µL of envenoming therapy. Curr. Pharm. Des. 22, 5270–5293 (2016).
the same buffer, twice with 200 µL of PBS, and finally twice with 200 µL of 6. Laustsen, A. H. Guiding recombinant antivenom development by omics
deionized water. After removing the supernatant from the last wash, the beads technologies. New Biotechnol. 45, 19-27 (2018).
were resuspended in 20 µL of a saturated solution of α-cyano-4-hydroxycinnamic 7. Harrison, R. A. & Gutiérrez, J. M. Priority actions and progress to substantially
acid in 50% acetonitrile and 50% water, containing 0.1% trifluoroacetic acid and 1 and sustainably reduce the mortality, morbidity and socioeconomic burden of
mg/mL of monobasic ammonium phosphate. The samples were mixed by vor- tropical snakebite. Toxins 8, 351 (2016).
texing for a few seconds, centrifuged for 15 s, and 1 µL of the supernatant was 8. Laustsen, A. H. et al. Exploration of immunoglobulin transcriptomes
spotted onto an Opti-TOF 384 plate, dried, and analyzed by MALDI-TOF on a from mice immunized with three-finger toxins and phospholipases A2 from the
4800-Plus Proteomics Analyzer (Applied Biosystems). TOF spectra were Central American coral snake, Micrurus nigrocinctus. PeerJ 5, e2924 (2017).
acquired in linear positive mode, using 500 shots at a laser intensity of 4200. 9. Tan, C. H., Tan, K. Y., Lim, S. E. & Tan, N. H. Venomics of the beaked sea
Identification of toxin family was achieved by comparing the obtained masses snake, Hydrophis schistosus: a minimalist toxin arsenal and its cross-
with previously reported values26. neutralization by heterologous antivenoms. J. Proteom. 126, 121–130 (2015).
10. Tan, N. H., Wong, K. Y. & Tan, C. H. Venomics of Naja sputatrix, the Javan
Animals. In vivo assays were conducted in CD-1 mice (18–20 g) of both sexes, spitting cobra: a short neurotoxin-driven venom needing improved antivenom
supplied by Instituto Clodomiro Picado, following protocols approved by the neutralization. J. Proteom. 157, 18–32 (2017).
Institutional Committee for the Use and Care of Animals (CICUA), University of 11. Laustsen, A. H., Lomonte, B., Lohse, B., Fernández, J. & Gutiérrez, J. M.
Costa Rica. Mice were housed in cages of various sizes for groups of 4–12, and were Unveiling the nature of black mamba (Dendroaspis polylepis) venom through
provided food and water ad libitum. venomics and antivenom immunoprofiling: identification of key toxin targets
for antivenom development. J. Proteom. 119, 126–142 (2015).
12. Harvey, A. L. Twenty years of dendrotoxins. Toxicon 39, 15–26 (2001).
Neutralization studies by intracerebroventricular injection. The neutralization 13. Laustsen, A. H., Johansen, K. H., Engmark, M. & Andersen, M. R.
activity of individual IgGs and IgG cocktails against both dendrotoxins and whole Recombinant snakebite antivenoms: a cost-competitive solution to a neglected
venom was tested by intracerebroventricular (i.c.v.) injection in groups of three to tropical disease? PLoS. Negl. Trop. Dis. 11, e0005361 (2017).
four mice (18–20 g body weight, both sexes used) using different doses of each 14. Roncolato, E. C. et al. Phage display as a novel promising antivenom therapy:
venom fraction and whole venom (0.5–1.5 µg per mouse) and different toxin:IgG
a review. Toxicon 93, 79–84 (2015).
molar ratios (1:0.5, 1:0.75, 1:1, 1:2, 1:3, 1:4, and 1:6). Venom and venom fraction
doses were selected as to ensure 100% mortality. IgGs and venom fractions/whole 15. Kitchens, C. & Eskin, T. Fatality in a case of envenomation by Crotalus
venom were mixed, and solutions were incubated (30 min either at room tem- adamanteus initially successfully treated with polyvalent ovine antivenom
perature or at 37 °C). After incubation, mice were injected with a volume of 5–33 followed by recurrence of defibrinogenation syndrome. J. Med. Toxicol. 4,
µL (dependent on IgG stock concentration). Irrelevant-specificity control mice 180–183 (2008).
were injected with venom fractions/whole venom and anti-lysozyme IgG dissolved 16. Laustsen, A. H. et al. Pros and cons of different therapeutic antibody formats
in phosphate-buffered saline (PBS; 0.12M NaCl, 0.04M sodium phosphate buffer, for recombinant antivenom development. Toxicon 146, 151–175 (2018).
pH 7.2), while vehicle-control mice were injected with venom fractions/whole 17. Calvete, J. J. & Lomonte, B. A bright future for integrative venomics. Toxicon
venom in PBS. Time of death was recorded, and Kaplan–Meier curves were used to 107, 159–162 Part B (2015).
represent mouse survival. For comparative purposes, the SAIMR polyvalent anti- 18. McCafferty, J., Griffiths, A. D., Winter, G. & Chiswell, D. J. Phage antibodies:
venom (South African Vaccine Producers, batch BC02645) was tested in a similar filamentous phage displaying antibody variable domains. Nature 348, 552–554
fashion. This antivenom is an F(ab’)2 preparation obtained from the plasma of (1990).
horses immunized with a mixture of the venoms of ten snake species, including D. 19. Martin, C. D. et al. A simple vector system to improve performance and
polylepis27. Various dilutions of this antivenom were incubated with a fixed con- utilisation of recombinant antibodies. BMC Biotechnol. 6, 46 (2006).
centration of D. polylepis venom. Controls included venom alone and antivenom 20. Studier, F. W. Protein production by auto-induction in high density shaking
alone. After incubation, aliquots of 10 µL, containing 1.5 µg venom, were injected i. cultures. Protein Expr. Purif. 41, 207–234 (2005).
c.v., as described, and deaths were recorded. In addition, antivenom neutralization 21. Laustsen, A. H., Lohse, B., Lomonte, B., Engmark, M. & Gutiérrez, J. M.
of lethality of D. polylepis venom was tested by the i.v. route, as previously Selecting key toxins for focused development of elapid snake antivenoms and
described11. inhibitors guided by a toxicity score. Toxicon 104, 43–45 (2015).
22. World Health Organisation. WHO Guidelines for the Production, Control and
Regulation of Snake Antivenom Immunoglobulins. Vol. 141 (World Health
Neutralization studies by intravenous injection. The neutralization activity of Organization, 2010).
the IgGs against three-finger toxins and whole venom was tested by intravenous
23. Laustsen, A. H., Lauridsen, L. P., Lomonte, B., Andersen, M. R. & Lohse, B.
(i.v.) injection in groups of four mice (18–20 g body weight), using a challenge dose
of 20.1 µg for venom fraction Dp6, 10.6 µg for venom fraction Dp7, and 25.8 µg of Pitfalls to avoid when using phage display for snake toxins. Toxicon 126,
whole venom, and a toxin:IgG molar ratio of 1:3. Venom and venom fraction doses 79–89 (2017).
were selected as to ensure 100% mortality. IgGs and venom fractions/whole venom 24. Schofield, D. J. et al. Application of phage display to high throughput antibody
were mixed and incubated (30 min at room temperature). Then, aliquots of the generation and characterization. Genome Biol. 8, R254 (2007).
mixtures were injected in the caudal vein, using an injection volume of 100–300 µL. 25. Pershad, K. et al. Generating a panel of highly specific antibodies to 20 human
Irrelevant-speci city control mice were injected with venom fractions/whole SH2 domains by phage display. Protein Eng. Des. Sel. 23, 279–288 (2010).fi
venom and anti-lysozyme IgG dissolved in PBS, while vehicle-control mice were
8 NATURE COMMUNICATIONS |          (2018)9 :3928 | DOI: 10.1038/s41467-018-06086-4 | www.nature.com/naturecommunications
NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06086-4 ARTICLE
26. Petras, D., Heiss, P., Harrison, R. A., Süssmuth, R. D. & Calvete, J. J. Top- Reprints and permission information is available online at http://npg.nature.com/
down venomics of the East African green mamba, Dendroaspis angusticeps, reprintsandpermissions/
and the black mamba, Dendroaspis polylepis, highlight the complexity of their
toxin arsenals. J. Proteom. 146, 148–164 (2016). Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in
27. Harrison, R. A. et al. Preclinical antivenom-efficacy testing reveals potentially published maps and institutional affiliations.
disturbing deficiencies of snakebite treatment capability in East Africa. PLoS
Negl. Trop. Dis. 11, e0005969 (2017).
Acknowledgements Open Access This article is licensed under a Creative Commons
We thank the following foundations and institutions for financial support: The Novo Attribution 4.0 International License, which permits use, sharing,
Nordisk Foundation (NNF16OC0019248), the Hørslev Foundation (203866), the Oticon adaptation, distribution and reproduction in any medium or format, as long as you give
Foundation (17-3802), IONTAS Ltd., and Instituto Clodomiro Picado. appropriate credit to the original author(s) and the source, provide a link to the Creative
Commons license, and indicate if changes were made. The images or other third party
Author contributions material in this article are included in the article’s Creative Commons license, unless
A.H.L., A.K.V., J.M.G., B.L. and J.M.C. conceived the study. A.H.L., A.K.V., E.W.M., indicated otherwise in a credit line to the material. If material is not included in the
A.S.A., U.P., C.K., S.O., P.S., D.T.G., A.M.L., R.A.L., M.L., J.M.G. and B.L. performed the article’s Creative Commons license and your intended use is not permitted by statutory
experiments. A.H.L., A.K.V., A.S.A., U.P., C.K., S.O., P.S., J.M.G., B.L. and J.M.C. ana- regulation or exceeds the permitted use, you will need to obtain permission directly from
lyzed the data. A.H.L. and A.K.V. wrote the manuscript. All authors revised and reviewed the copyright holder. To view a copy of this license, visit http://creativecommons.org/
the manuscript. licenses/by/4.0/.
Additional information © The Author(s) 2018
Competing interests: The authors declare no competing interests.
NATURE COMMUNICATIONS |      (    2018) 9:3928 | DOI: 10.1038/s41467-018-06086-4 | www.nature.com/naturecommunications 9