N-terminal domain on dystroglycan enables LARGE1 to extend matriglycan on α-dystroglycan and prevents muscular dystrophy

  1. Hidehiko Okuma
  2. Jeffrey M Hord
  3. Ishita Chandel
  4. David Venzke
  5. Mary E Anderson
  6. Ameya S Walimbe
  7. Soumya Joseph
  8. Zeita Gastel
  9. Yuji Hara
  10. Fumiaki Saito
  11. Kiichiro Matsumura
  12. Kevin P Campbell  Is a corresponding author
  1. Howard Hughes Medical Institute, Senator Paul D. Wellstone Muscular Dystrophy Specialized Research Center, Department of Molecular Physiology and Biophysics and Department of Neurology, Roy J. and Lucille A. Carver College of Medicine, The University of Iowa, United States
  2. Department Pharmaceutical Sciences, School of Pharmaceutical Sciences, University of Shizuoka, Japan
  3. Department of Neurology, School of Medicine, Teikyo University, Japan
7 figures, 1 table and 1 additional file

Figures

Domain structure of dystroglycan (DG) and Δ-α-DGN.

Wild-type DG is a pre-proprotein with an N-terminal signal peptide (light green) that is translated in the rough endoplasmic reticulum. The globular N-terminal domain (α-DGN; orange) is present in wild-type DG but absent in the mutant (∆-α-DGN). The junction between α-DGN and the mucin-like domain (light teal) contains a furin convertase site. The globular extracellular C-terminal domain (CTD; pink) contains an SEA (sea urchin sperm protein, enterokinase and agrin) autoproteolysis site, which cleaves pro-DG into α-DG and β-DG (green). Glycosylation has been omitted for clarity.

Figure 2 with 2 supplements
Characterization of mice with a muscle-specific loss of α-DG N-terminal (α-DGN).

(A) Body weight and grip strength of 12-week-old wild-type (WT) littermate (control) and muscle-specific α-DGN knockout (M-α-DGN KO) mice. Double and quadruple asterisks: statistical significance determined by Student’s unpaired t-test (**p-value = 0.005, ****p-value <0.0001). (B) Histological analyses of quadriceps muscles from 12-week-old control and M-α-DGN KO mice. Sections stained with H&E or used for immunofluorescence to detect β-DG (affinity purified rabbit anti-β-DG), DAPI, and matriglycan (IIH6). Scale bars = 50 μm. (C) Immunoblot analysis of skeletal muscle from control, M-α-DGN KO, and Large1myd mice. Glycoproteins were enriched using wheat-germ agglutinin (WGA)-agarose with 10 mM EDTA. Immunoblotting was performed to detect matriglycan (IIIH11), core α-DG, β-DG (AF6868), and laminin overlay. α-DG in WT control muscle (α-DG (WT)) and α-DG in α-DGN-deficient muscle (α-DG (Δα-DGN)) are indicated on the right. The number of KO mice was 17, and that of LC mice was 57. There were 6 male KO mice, 11 female KO mice, 23 male LC mice, and 34 female LC mice. Molecular weight standards in kilodaltons (kDa) are shown on the left.

Figure 2—figure supplement 1
Mice heterozygous (+/-) for a constitutive deletion of α-DG N-terminal (α-DGN) have two different sizes of α-DG.

Immunoblot analysis of skeletal muscle from littermate controls or mice that are heterozygous for the α-DGN KO allele (α-DGN (-/+)). Glycoproteins were enriched from the quadriceps skeletal muscles of mice using wheat-germ agglutinin (WGA)-agarose with 10 mM EDTA. Immunoblotting was performed to detect matriglycan (IIIH11), core α-DG and β-DG (AF6868), and laminin overlay. α-DG in wild-type (WT) control muscle (α-DG(WT)) and α-DG in α-DGN-deficient muscle (α-DG(Δα-DGN)) are indicated on the right. Molecular weight standards in kilodaltons (kDa) are shown on the left.

Figure 2—figure supplement 2
The short 100–120 kDa band in muscle-specific α-DGN knockout (M-α-DGN KO) muscle is matriglycan.

(A) Immunoblot analysis of total skeletal muscle from control mice after digestion with enzymes β-glucuronidase and α-xylosidase. Glycoproteins were enriched using wheat-germ agglutinin (WGA)-agarose with 10 mM EDTA and incubated overnight with β-glucuronidase (BGUS) and α-xylosidase (XyIS). Immunoblotting was performed to detect matriglycan (IIH6), core α-DG and β-DG (AF6868), and laminin overlay before (To) and after overnight digestion (TO/N). (B) Immunoblot analysis of M-α-DGN KO total skeletal muscle after digestion with enzymes BGUS and XyIS. Glycoproteins were enriched using WGA-agarose with 10 mM EDTA and incubated overnight with BGUS and XyIS. Immunoblotting was performed to detect matriglycan (IIH6), core α-DG and β-DG (AF6868), and laminin overlay before (To) and after digestion (TO/N). Molecular weight standards in kilodaltons (kDa) are shown on the left.

α-DG N-terminal (α-DGN) deficiency results in post-synaptic defects.

Neuromuscular junctions (NMJs) from tibialis anterior (TA), extensor digitorum longus (EDL), and soleus (SOL) muscles obtained from 35- to 39-week-old control and muscle-specific α-DGN knockout (M-α-DGN KO) mice. (A) Representative images of post-synaptic terminals (α-BTX-488; green), motor axons (anti-neurofilament-H; red), and pre-synaptic terminals (anti-synaptophysin; red) from TA, EDL, and SOL muscles. Scale bars = 20 μm. (B) Scoring of post-synaptic defects by blinded observers (scoring criteria described in Materials and methods). Statistical significance determined by Student’s unpaired t-test; *p-value <0.05; **p-value <0.001; ***p-value <0.0001.

Figure 4 with 1 supplement
α-DG N-terminal (α-DGN)-deficient extensor digitorum longus (EDL) muscle demonstrates greater lengthening contraction-induced force decline.

(A) Weight (mg) of EDL muscles from wild-type (WT) littermates (controls) and muscle-specific α-DGN knockout (M-α-DGN KO) mice; p=0.2469, as determined by Student’s unpaired t-test. (B) Cross-sectional area of EDL muscles; p=0.1810, as determined by Student’s unpaired t-test. (C) Maximum absolute tetanic force production in EDL muscles. p=0.0488, as determined by Student’s unpaired t-test. (D) Specific force production in EDL muscles; p=0.4158, as determined by Student’s unpaired t-test. (E) Force deficit and force recovery after eccentric contractions in EDL muscles from 12- to 17-week-old male and female control (closed circles; n=7) and M-α-DGN KO (open circles; n=7) mice. *p<0.05; **p<0.01; ***p<0.001, as determined by Student’s unpaired t-test of at any given lengthening contractions cycle. Bars represent the mean ± the standard deviation.

Figure 4—figure supplement 1
α-DG N-terminal (α-DGN)-deficient muscle and protein O-mannose kinase (POMK)-deficient muscle with similar short forms of matriglycan exhibit similar lengthening contraction-induced force decline.

Force deficit and force recovery after eccentric contractions in extensor digitorum longus (EDL) muscles from 12- to 17-week-old male and female controls (closed circles; n=7), muscle-specific α-DGN knockout (M-α-DGN KO) (open circles; n=7), M-POMK littermate controls (closed triangles; n=3), and M-POMK KO (open triangles; n=4) mice. There is no significant difference in M-α-DGN KO vs. M-POMK KO EDL as determined by Student’s unpaired t-test at any given lengthening contractions cycle and post-lengthening contraction.

Figure 5 with 2 supplements
Exogenous α-DG N-terminal (α-DGN)-deficient dystroglycan (DG) also produces short matriglycan like M-Dag1 KO muscle.

(A) Schematic representation of a wild-type (WT) DG and an adeno-associated virus (AAV) carrying a mutant DG in which the N-terminal domain has been deleted (DG-E). α-DG is composed of a signal peptide (SP, amino acids 1–29), an N-terminal domain (amino acids 30–316), a mucin-like domain (amino acids 317–485), and a C-terminal domain (amino acids 486–653). The green box represents β-DG. (B) Immunofluorescence analyses of quadriceps muscles from 12-week-old M-Dag1 KO mice injected with AAV-MCK DG-E to detect β-DG, nuclei (DAPI), and matriglycan (IIH6). Scale bars = 50 μm. (C) Immunoblot analysis of skeletal muscle obtained from littermate controls (control), M-Dag1 KO mice, or M-Dag1 KO mice injected with AAV-MCK DG-E. Glycoproteins were enriched from skeletal muscles using wheat-germ agglutinin (WGA)-agarose. Immunoblotting was performed to detect matriglycan (IIIH11), core α-DG and β-DG (AF6868), and laminin (overlay). (D) Production of specific force in extensor digitorum longus ( EDL) muscles from 12- to 17-week-old male and female M-Dag1 KO mice (controls; closed circles, n=10); M-α-DGN KO mice (open circles, n=13); and M-Dag1 KO + AAV MCK DG-E mice (open triangles, n=6). p-Values determined by Student’s unpaired t-test; controls vs. M-Dag1 KO: p=0.4158; controls vs. M-Dag1 KO + AAV MCK DG-E: p=0.3632; M-Dag1 KO vs. M-Dag1 KO + AAV MCK DG-E: p=0.948. (E) Force deficits and recovery in EDL muscles from mice in D. There is no significant difference in M-Dag1 KO vs. M-Dag1 KO + AAV MCK DG-E as determined by Student’s unpaired t-test at any given lengthening contraction cycle or post-lengthening contraction.

Figure 5—figure supplement 1
Characteristics of M-Dag1 KO (Pax7cre; Dag1flox/flox) mice.

(A) Immunofluorescence analyses of quadriceps muscles from a 12-week-old wild-type (WT) littermate (control) or M-Dag1 KO mouse. Sections were stained to detect β-DG (AP83) and nuclei (DAPI). Scale bars = 50 μm. (B) Immunoblot analysis of skeletal muscle from control and M-Dag1 KO mice. Glycoproteins were enriched from skeletal muscles using wheat-germ agglutinin (WGA)-agarose with (+) and without (-) 10 mM EDTA. Immunoblotting was performed to detect matriglycan (IIIH11) and core α-DG and β-DG (AF6868). (C) Specific force in extensor digitorum longus (EDL) muscles of mice in indicated groups; p=0.0128, as determined by Student’s unpaired t-test. (D) Force deficit and force recovery after eccentric contractions in EDL muscles of 12- to 17-week-old male and female control (n=3) and M-Dag1 KO (n=6) mice.

Figure 5—figure supplement 2
Excess free α-DG N-terminal (α-DGN) interferes with like-acetylglucosaminyltransferase-1 (LARGE1) elongation of matriglycan on α-DG.

Immunoblot analysis of skeletal muscle from control and α-DGN transgenic (α-DGN Tg) mice. Glycoproteins were enriched using wheat-germ agglutinin (WGA)-agarose. Immunoblotting was performed to detect matriglycan (IIH6), core α-DG, and β-DG (AF6868).

Figure 6 with 1 supplement
Expression of α-DG N-terminal (α-DGN) in muscle-specific α-DGN knockout (M-α-DGN KO) mice does not rescue matriglycan elongation.

(A) Representative sections of quadriceps muscles from 17-week-old M-α-DGN KO mice injected with AAV-CMV α-DGN. Sections were stained with H&E and immunofluorescence to detect matriglycan (IIH6) and β-DG (AP83). Scale bars = 50 μm.(B) Immunoblot analysis of skeletal muscle obtained from littermate controls or M-α-DGN KO mice and M-α-DGN KO mice injected with AAV-CMV α-DGN (M-α-DGN KO+AAV CMV α-DGN). Glycoproteins were enriched using wheat-germ agglutinin (WGA)-agarose with 10 mM EDTA. Immunoblotting was performed to detect matriglycan (IIIH11), core α-DG and β-DG (AF6868), and laminin overlay. (C) Production of specific force in extensor digitorum longus (EDL) muscles from 12- to 17-week-old male and female M-α-DGN wild-type (WT) littermates (controls; closed circles, n=10); M-α-DGN KO (open circles, n=13); and M-α-DGN KO+AAV CMV α-DGN (closed triangles, n=12). p-Values determined by Student’s unpaired t-test; controls vs. M-α-DGN KO+AAV CMV α-DGN: p=0.8759; controls vs. M-α-DGN KO: p=0.4333; M-α-DGN KO vs. M-α-DGN KO+AAV CMV α-DGN: p=0.4333. (D) Force deficit and force recovery after lengthening contractions in EDL muscles from 12- to 17-week-old male and female M-α-DGN KO WT littermates (controls, closed circles; n=6) and M-α-DGN KO (KO, open circles; n=7) mice, and in M-α-DGN KO mice injected with AAV-CMV α-DGN (KO+AAV CMV α-DGN, closed triangles; n=8). There is no significant difference in M-α-DGN KO vs. M-α-DGN KO+AAV CMV α-DGN as determined by Student’s unpaired t-test at any given lengthening contractions cycle or post-lengthening contractions.

Figure 6—figure supplement 1
Like-acetylglucosaminyltransferase-1 (LARGE1) overexpression does not extend matriglycan on dystroglycan (DG) lacking α-DG N-terminal (α-DGN).

AAV-MCK-Large1 was injected into the retro-orbital sinus 10- to-24-week-old muscle-specific α-DGN knockout (M-α-DGN KO) mice. Quadriceps skeletal muscle was dissected 10–22 weeks after injection from control, M-α-DGN KO, and M-α-DGN KO+AAV-MCK-mLarge1 and used for immunoblotting analysis. Glycoproteins were enriched using wheat-germ agglutinin (WGA)-agarose with 10 mM EDTA. Immunoblotting was performed to detect matriglycan (IIIH11), core α-DG and β-DG (AF6868), and laminin (overlay). Molecular weight standards in kilodaltons (kDa) are shown on the left.

Relationship between matriglycan length and dystrophic phenotype.

(A) Percentage deficit of eight eccentric contraction (EC) in extensor digitorum longus (EDL) muscles from C57BL/6J wild-type (WT) (control), muscle-specific α-DGN knockout (M-α-DGN KO), and Dag1T190M mice. p-Values determined by Student’s unpaired t-test; control vs. Dag1T190M: p=0.0263; control and Dag1T190M vs. M-α-DGN KO: p<0.001. (B) Immunoblot analysis of quadriceps skeletal muscles from control, Dag1T190M and M-α-DGN KO mice. Glycoproteins were enriched using wheat-germ agglutinin (WGA)-agarose with 10 mM EDTA. Immunoblotting was performed with laminin (laminin overlay). (C) Percentage of muscle fibers with central nuclei in 12- to 19-week-old control, Dag1T190M and M-α-DGN KO mice; n=4 for all groups. p-Values determined by Student’s unpaired t-test; control and Dag1T190M vs. M-α-DGN KO: p<0.001; control vs. Dag1T190M: p=0.0263. (D) Comparison of average solid-phase determined relative Bmax values for laminin. Bmax values for M-α-DGN KO were set to 1 to allow for direct comparisons; error bars indicate s.e.m. p-Values determined using Student’s unpaired t-test; control vs. Dag1T190M and control vs. M-α-DGN KO: p<0.01, and M-α-DGN KO vs. Large1myd: p<0.001. (E) Solid-phase analysis of laminin-binding using laminin-111 in skeletal muscle from control, Dag1T190M, M-α-DGN KO, and Large1myd KO mice (three replicates for each group). Control Kd: 0.9664±0.06897 nM; Dag1T190M Kd: 1.902±0.1994 nM; and M-α-DGN KO Kd: 2.322±0.6114 nM.

Tables

Appendix 1—key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Genetic reagent
(Mus musculus)
Pax7Cre C57BL/6JThe Jackson Laboratory, Bar Harbor, ME, USAJAX:010530,
RRID:IMSR_JAX:010530
Pax7tm1(cre)Mrc
Genetic reagent
(Mus musculus)
LargemydThe Jackson Laboratory, Bar Harbor, ME, USAJAX:000300, RRID:IMSR_JAX:000300MYD/Le-Os+/+Largemydmyd/J
Genetic reagent
(Mus musculus)
LargemydCampbell LabDescribed in Materials and Methods: Animals
(92.5% C57BL/6J)
Genetic reagent
(Mus musculus)
Dag1ΔH30-A316PMID:31097590
DOI:10.1073/pnas.1904493116
Dag1Δα-DGN
Genetic reagent
(Mus musculus)
Dag1floxPMID:12230980
DOI: 10.1016/s0092-8674(02)00907–8
JAX:006834, RRID:IMSR_JAX:006834B6.129-Dag1tm2Kcam/J
Genetic reagent
(Mus musculus)
Dag1T190MPMID: 21388311
DOI: 10.1056/NEJMoa1006939
Genetic reagent
(Mus musculus)
α-DGN TgThis paperDescribed in Materials and methods: Animals
Genetic reagent
(Mus musculus)
PomkfloxPMID:32975514
DOI:10.7554/eLife.61388
AntibodyAnti-DG; sheep polyclonalR&D SystemsCat# AF6868, RRID:AB_10891298WB (1:500)
AntibodyAnti-α-DG (IIH6C4); mouse monoclonalDevelopment Studies Hybridoma Bank/Campbell LabCat# IIH6 C4, RRID:AB_2617216Described in Materials and Methods: Animals
IF (1:10-1:100)
AntibodyAnti-α-DG (IIH6C4); mouse monoclonalMilliporeSigma
Campbell Lab
Cat# 05–593, RRID:AB_309828Described in Materials and Methods: Animals

WB (1:1000–1:2000)
AntibodyAnti-Laminin; rabbit polyclonalMilliporeSigmaCat# L9393, RRID:AB_477163WB (1:1000), Solid Phase Assay (1:5000)
AntibodyAnti-β-DG; rabbit polyclonalCampbell Lab
PMID: 1741056
DOI: 10.1038/355696a0
AP83Described in Materials and Methods: Animals

IF (1:50)
AntibodyAnti-β-DG mouse IgM; mouse monoclonalLeica BiosystemsCat# NCL-b-DG, RRID:AB_442043IF (1:50 to 1:200)
AntibodyAnti-sheep IgG; donkey polyclonalRocklandCat# 613-731-168, RRID:AB_220181WB (1:2000)
AntibodyAnti-mouse IgG (H+L); donkey polyclonalLI-COR BiosciencesCat# 926–32212, RRID:AB_621847WB (1:15,000), IF (1:800)
AntibodyAnti-rabbit IgG (H+L); donkey polyclonalLI-COR BiosciencesCat# 926–32213, RRID:AB_621848WB (1:15,000), IF (1:800)
AntibodyAnti-mouse IgM; goat polyclonalLI-COR BiosciencesCat# 926–32280, RRID:AB_2814919WB (1:2500)
AntibodyAnti-mouse IgG1; goat polyclonalLI-COR BiosciencesCat# 926–32350, RRID:AB_2782997WB (1:2000, 1:10,000)
AntibodyAnti-rabbit IgG (H+L); goat polyclonalThermo Fisher ScientificCat# A-11034, RRID:AB_2576217IF (1:1000 to 1:2000)
AntibodyAnti-mouse IgM; goat polyclonalThermo Fisher ScientificCat# A-21042, RRID:AB_2535711IF (1:1000 to 1:2000)
AntibodyAnti-human Synaptophysin (SP11); rabbit monoclonalThermo Fisher ScientificCat# MA5-14532, RRID:AB_10983675IF (1:100)
AntibodyNeurofilament NF-H; chicken polyclonalEnCor BiotechnologyCat# CPCA-NF-H, RRID:AB_2149761IF (1:1000)
AntibodyAnti-chicken IgY (H+L); goat polyclonalThermo Fisher ScientificCat# A32759, RRID:AB_2762829IF (1:1000)
AntibodyAnti-α-DG (IIIH11); mouse monoclonalCampbell LabDescribed in Materials and Methods: Animals

WB (1:100–1:1000)
Chemical compound, drugPepstatin AMilliporeSigmaCat# 516481
Chemical compound, drugCalpain Inhibitor IMilliporeSigmaCat# A6185
Chemical compound, drugAprotinin from bovine lungMilliporeSigmaCat# A1153
Chemical compound, drugLeupeptinMilliporeSigmaCat# 108975
Chemical compound, drugPMSFMilliporeSigmaCat# P7626
Chemical compound, drugImmobilon-FL PVDFMilliporeSigmaCat# IPFL00010
Chemical compound, drugCalpeptinThermo Fisher ScientificCat# 03-340-05125M
Chemical compound, drugBis-acrylamide solution-30% (37.5:1)Hoefer, IncCat# GR337-500
Chemical compound, drugBenzamidine Hydrochloride HydrateMP BiomedicalsCat# 195068
Chemical compound, drugWGA agarose boundVector LabsCat# AL-1023, RRID:AB_2336862
Chemical compound, drugPrecision Plus Protein All Blue Standards-500 µLBio-RadCat# 161–0373
Chemical compound, drugEthylenediamine Tetraacetic acid, disodium salt dihydrate, EDTAThermo Fisher ScientificCat# S311-500
Peptide, recombinant proteinEnzymes,
β-glucuronidase
α-xylosidase
This paper and PMID: 27526028
DOI: 10.1038/nchembio.2146
Described in Materials and methods: Digestion of α-DG with exoglycosidases
Software, algorithmSigmaPlotSigmaPlotRRID:SCR_003210
Software, algorithmExcelMicrosoftRRID:SCR_016137
Software, algorithmGraphPad PrismGraphPadRRID:SCR_002798Version 8.3
Software, algorithmFlowJoBecton, Dickinson & Company (BD)RRID:SCR_008520Version 7.6.5
Software, algorithmImage Studio Acquisition SoftwareLI-COR BiosciencesRRID:SCR_015795
Software, algorithmFijiNational Institutes of HealthRRID:SCR_002285
SoftwareAdobe IllustratorAdobeRRID:SCR_010279Version 27.1.1
Software, algorithmUniProt ProteomesRRID:SCR_018666
Software, algorithmIUPREDRRID:SCR_014632
OtherStreptavidin, Alexa Fluor 594 conjugateThermo Fisher ScientificCat# S11227IF (1:1000 to 1:2000)
OtherWestern Blot ImagerLI-COR BiosciencesOdyssey CLx
RRID:SCR_014579
OtherIsolated Mouse Muscle SystemAurora Scientific1200A
OtherMouse Grip Strength MeterColumbus Instruments1027 Mouse
OtherTabletop ultracentrifugeBeckman CoulterOptima MAX, 130K
OtherUltracentrifugeBeckman CoulterOptima-L-100 XP
OtherCentrifugeBeckman CoulterAvanti J-E HPC
OtherSlide Scanner MicroscopeOlympusVS120-S5-FL
RRID:SCR_018411
OtherConfocal MicroscopeOlympusFLUOVIEW FV3000
RRID:SCR_017015
OtherCryostatLeica BiosystemsCM3050S
RRID:SCR_016844
OtherImaging SystemLI-COR BiosciencesOdyssey CLx Infrared
RRID:SCR_014579
Other96-Well PlatesCorning IncCat# 3590Described in Materials and methods
OtherHematoxylin (Certified Biological Stain)Fisher ScientificCat# H345-100Described in Materials and methods
OtherEosin 515 LTLeica BiosystemsCat# 3801619Described in Materials and methods
OtherAAV-MCK DG (ΔH30-A316)This paperDescribed in Materials and methods: AAV vector production and AAV injection
OtherAAV-CMV-α-DGNThis paperDescribed in Materials and methods: AAV vector production and AAV injection
OtherAAV-MCK-mLarge1This paperDescribed in Materials and methods: AAV vector production and AAV injection
Otherα BTX-488
(α-Bungarotoxin)
Thermo Fisher ScientificCat# B13423IF (1:500)
OtherLaminin (Natural, mouse) Lam-111GibcoCat# 23017–015
OtherEnzymes,
β-glucuronidase
α-xylosidase
This paper and PMID: 27526028
DOI: 10.1038/nchembio.2146
Described in Materials and methods: Digestion of α-DG with exoglycosidases

Additional files

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Hidehiko Okuma
  2. Jeffrey M Hord
  3. Ishita Chandel
  4. David Venzke
  5. Mary E Anderson
  6. Ameya S Walimbe
  7. Soumya Joseph
  8. Zeita Gastel
  9. Yuji Hara
  10. Fumiaki Saito
  11. Kiichiro Matsumura
  12. Kevin P Campbell
(2023)
N-terminal domain on dystroglycan enables LARGE1 to extend matriglycan on α-dystroglycan and prevents muscular dystrophy
eLife 12:e82811.
https://doi.org/10.7554/eLife.82811