1. Cell Biology
  2. Human Biology and Medicine
Download icon

Accelerated phosphatidylcholine turnover in macrophages promotes adipose tissue inflammation in obesity

  1. Kasparas Petkevicius  Is a corresponding author
  2. Sam Virtue
  3. Guillaume Bidault
  4. Benjamin Jenkins
  5. Cankut Çubuk
  6. Cecilia Morgantini
  7. Myriam Aouadi
  8. Joaquin Dopazo
  9. Mireille J Serlie
  10. Albert Koulman
  11. Antonio Vidal-Puig  Is a corresponding author
  1. University of Cambridge Metabolic Research Laboratories, Institute of Metabolic Science, MDU MRC, United Kingdom
  2. Fundación Progreso y Salud, CDCA, Hospital Virgen del Rocio, Spain
  3. INB-ELIXIR-es, FPS, Hospital Virgen del Rocio, Spain
  4. Centro de Investigación Biomédica en Red de Enfermedades Raras (CIBERER), FPS, Hospital Virgen del Rocio, Spain
  5. Karolinska Institutet, Sweden
  6. Amsterdam University Medical Center, Netherlands
  7. Wellcome Trust Sanger Institute, United Kingdom
Research Article
Cite this article as: eLife 2019;8:e47990 doi: 10.7554/eLife.47990
7 figures, 1 table, 2 data sets and 4 additional files

Figures

Figure 1 with 5 supplements
De novo PC synthesis rate is increased in ATMs during obesity.

(A) Simplified schema of the Kennedy pathway of de novo PC biosynthesis. Transcripts in red are upregulated in 16-week-old Lepob/ob eWAT macrophages compared to WT controls. (B) Normalised microarray gene expression values for the enzymes of the Kennedy pathway of de novo PC biosynthesis in eWAT macrophages. (C) De novo PC biosynthesis pathway module (M00090) activity in eWAT macrophages, as inferred by the Metabolizer algorithm. (D) Gene Ontology (GO) pathways related to membrane lipid metabolism that are increased in Lepob/ob eWAT macrophages at 16 weeks, but not at 5 weeks of age compared to WT controls. (EPcyt1a expression in eWAT macrophages, measured by qPCR. (F)Pcyt1a expression, measured by qPCR in subcutaneous WAT macrophages isolated from obese patients undergoing bariatric surgery, plotted against their body weight (n = 19). (G) Molar abundance of 16:0 and 18:0 lyso-PC species (expressed as percentage of total measured PLs) in eWAT macrophages (n = 3 pools of 5 WT, n = 6 Lepob/ob mice). ∗p<0.05 between genotypes, error bars indicate SEM.

https://doi.org/10.7554/eLife.47990.003
Figure 1—figure supplement 1
Obesity does not affect de novo PE synthesis rate in ATMs.

(A) Normalised microarray gene expression for the enzymes of the Kennedy pathway of de novo PE biosynthesis in eWAT macrophages. (B) De novo PE biosynthesis pathway module (M00092) activity in eWAT macrophages, as inferred by the Metabolizer algorithm. (CPcyt1b expression in eWAT macrophages, measured by qPCR.

https://doi.org/10.7554/eLife.47990.004
Figure 1—figure supplement 2
Obesity does not affect de novo PC synthesis rate in liver macrophages.

Normalised RNAseq gene expression for the enzymes of the Kennedy pathway of de novo PC biosynthesis in liver macrophages, isolated from 14-week-old WT and Lepob/ob mice (n = 3).

https://doi.org/10.7554/eLife.47990.005
Figure 1—figure supplement 3
Pcyt1a expression levels in different tissue macrophage populations.

Figure indicating Pcyt1a mRNA levels in isolated tissue macrophage populations, measured by RNAseq was obtained directly from Immgen database (www.immgen.org).

https://doi.org/10.7554/eLife.47990.006
Figure 1—figure supplement 4
Obesity tends to increase PC molar percentage and PC to PE molar ratio in ATMs.

(A) PC molar abundance (expressed as percentage of total measured PLs), and (B) PC:PE molar ratio in eWAT macrophages (n = 3 pools of 5 WT, n = 6 16 week-old Lepob/ob mice).

https://doi.org/10.7554/eLife.47990.007
Figure 1—figure supplement 5
The effects of obesity on PC remodelling gene expression in ATMs.

(A) Normalised microarray gene expression for the enzymes involved in PC remodelling in eWAT macrophages. (B) Simplified schema of the interaction between de novo PC synthesis and PC remodelling pathways.

https://doi.org/10.7554/eLife.47990.008
Figure 2 with 4 supplements
Myeloid cell-specific deletion of Pcyt1a leads to improved systemic glucose metabolism on the Lepob/ob genetic background.

(A) Schema of the BMT study design. (B) Body weight gain curves and (C) weights of indicated tissues of Lepob/ob mice transplanted with Pcyt1afl/fl (n = 7) or Pcyt1afl/fl Lyz2Cre/+ (n = 8) bone marrow. (D)GTT curves and areas under curve (AUC), normalised to basal glucose levels. (E) ITT curves, presented as percentage values of basal glucose levels, and areas above curve (AAC), normalised to basal glucose levels.

https://doi.org/10.7554/eLife.47990.009
Figure 2—figure supplement 1
Myeloid cell-specific deletion of Pcyt1a does not impair BMDM differentiation or function in vitro.

(A) Flow cytometry quantification of macrophage differentiation markers. (B) Phagocytosis of E. coli (normalised to control values). (C) Cytokine secretion into the medium of BMDMs stimulated with LPS for 6 or 24 hr. (D) Pcyt1a expression, CCTα protein levels and 3H-choline incorporation rate into membrane lipids. Pcyt1afl/fl (n = 4) or Pcyt1afl/fl Lyz2Cre/+ (n = 4) BMDMs.

https://doi.org/10.7554/eLife.47990.010
Figure 2—figure supplement 2
Myeloid cell-specific deletion of Pcyt1a does not affect eWAT or liver gene expression in lean mice.

Normalised expression of indicated genes in (A) eWAT and (B) liver of Pcyt1afl/fl (n = 7) or Pcyt1afl/fl Lyz2Cre/+ (n = 9) animals on a C57Bl/6J genetic background, measured by qPCR.

https://doi.org/10.7554/eLife.47990.011
Figure 2—figure supplement 3
Myeloid cell-specific deletion of Pcyt1a does not affect growth or glucose metabolism of lean mice.

(A) Body weight gain curves and (B) tissue weights of lean mice. (C) GTT curves and areas under curve (AUC), normalised to basal glucose levels. (D) ITT curves, presented as percentage values of basal glucose levels, and areas above curve (AAC), normalised to basal glucose levels.

https://doi.org/10.7554/eLife.47990.012
Figure 2—figure supplement 4
WT to Lepob/ob bone marrow transplant does not affect the increase of Pcyt1a transcription in eWAT ATMs.

Pcyt1a expression measured by qPCR in eWAT macrophages isolated from WT and Lepob/ob mice carrying WT bone marrow at 18 weeks of age (n = 7 mice/group).

https://doi.org/10.7554/eLife.47990.013
Figure 3 with 3 supplements
Myeloid cell-specific deletion of Pcyt1a leads to reduced WAT inflammation on the Lepob/ob genetic background.

(A) Selected pathways from eWAT RNAseq analysis, upregulated (red) or downregulated (blue) in Lepob/ob mice transplanted with Pcyt1afl/fl Lyz2Cre/+ (n = 8) bone marrow compared to controls (n = 7). (B) Representative AKT phosphorylation Western blots and (C) their densitometry quantification in eWAT of Lepob/ob mice transplanted with Pcyt1afl/fl (n = 7) or Pcyt1afl/fl Lyz2Cre/+ (n = 8) bone marrow. (D) Flow cytometry gating strategy, quantification of ATM number per gram of eWAT and the relative polarisation of ATM population in Lepob/ob mice.

https://doi.org/10.7554/eLife.47990.014
Figure 3—figure supplement 1
The effects of myeloid cell-specific deletion of Pcyt1a on eWAT and liver gene expression on the Lepob/ob genetic background.

Relative expression of indicated genes in the (A) eWAT and (B) liver of Lepob/ob BMT mice, measured by qPCR.

https://doi.org/10.7554/eLife.47990.015
Figure 3—figure supplement 2
Myeloid cell-specific deletion of Pcyt1a does not affect eWAT CLS number or adipocyte size on the Lepob/ob genetic background.

(A) Representative histology images of eWAT sections of Lepob/ob BMT mice, CLS are marked in red. Quantification of (B) CLS and (C) average adipocyte area in the eWAT sections of Lepob/ob BMT mice.

https://doi.org/10.7554/eLife.47990.016
Figure 3—figure supplement 3
The effects of myeloid cell-specific deletion of Pcyt1a on liver and skeletal muscle AKT signalling on the Lepob/ob genetic background.

Representative AKT phosphorylation Western blots and their densitometry quantification in (A, B) liver and (C, D) gastrocnemius muscle of Lepob/ob mice transplanted with Pcyt1afl/fl (n = 7) or Pcyt1afl/fl Lyz2Cre/+ (n = 8) bone marrow.

https://doi.org/10.7554/eLife.47990.017
Figure 4 with 2 supplements
Pcyt1a deficiency protects macrophages from palmitate-induced ER stress and inflammation.

(ATnf expression levels in Pcyt1afl/fl (n=5) or Pcyt1afl/fl Lyz2Cre/+ (n=3) BMDMs treated with indicated doses of palmitate for 16 hours. (B) Fold induction (compared to BSA alone) of indicated ER stress marker gene expression Pcyt1afl/fl (n=7) or Pcyt1afl/fl Lyz2Cre/+ (n=8) BMDMs treated with 250 μM palmitate for 16 hours. (C) Representative Western blots and (D) their densitometry quantification of Pcyt1afl/fl (n=5) or Pcyt1afl/fl Lyz2Cre/+ (n=3) BMDMs treated with indicated doses of palmitate for 16 hours. ∗p < 0.05 between genotypes, error bars indicate SEM. All presented experiments are representative of at least 3 BMDM cultures.

https://doi.org/10.7554/eLife.47990.018
Figure 4—figure supplement 1
Pcyt1a deficiency protects macrophages from palmitate, but not thapsigargin-induced cell death.

Flow cytometry quantification of dead Pcyt1afl/fl (n = 4) or Pcyt1afl/fl Lyz2Cre/+ (n = 4) BMDMs treated with indicated doses of (A) palmitate or (B) 150 nM thapsigargin for 16 hr.

https://doi.org/10.7554/eLife.47990.019
Figure 4—figure supplement 2
Pcyt1a deficiency protects peritoneal macrophages from palmitate-induced ER stress.

ER stress marker gene expression in cultured peritoneal macrophages from Pcyt1afl/fl (n = 4) or Pcyt1afl/fl Lyz2Cre/+ (n = 8) treated with 250 μM palmitate for 16 hr.

https://doi.org/10.7554/eLife.47990.020
Pcyt1a deficiency in macrophages does not affect the rate of exogenous palmitate incorporation into PCs.

(A) Representative autoradiogram and densitometry quantification of 14C-palmitate incorporation into PCs or (B) total lipids of Pcyt1afl/fl (n = 4) or Pcyt1afl/fl Lyz2Cre/+ (n = 4) BMDMs treated with 250 μM palmitate for indicated periods of time. (C) Inhibition of de novo PC biosynthesis (red line) and 14C-palmitate incorporation into PC fraction (blue line) of WT BMDMs (n = 4), pretreated with indicated doses of miltefosine for 1 hr and stimulated with 250 μM palmitate for 3 hr. Representative autoradiogram is presented below. (D) Representative autoradiogram and densitometry quantification of 14C-acetate incorporation into PCs or (E) total lipids of untreated Pcyt1afl/fl (n = 4) or Pcyt1afl/fl Lyz2Cre/+ (n = 4) BMDMs over 3 hr, normalised to Pcyt1afl/fl group average. ∗p<0.05 between genotypes, error bars indicate SEM. All presented experiments are representative of at least 3 BMDM cultures.

https://doi.org/10.7554/eLife.47990.021
Figure 6 with 4 supplements
Pcyt1a deficiency increases PUFA-containing PC levels in macrophages.

(A) Volcano plot of indicated PC species, expressed as molar percentage of all measured PCs, in Pcyt1afl/fl (n = 7) or Pcyt1afl/fl Lyz2Cre/+ (n = 8) BMDMs. (B) Pie charts indicating the relative abundance of PC species with different degrees of unsaturation in Pcyt1afl/fl (n = 7) or Pcyt1afl/fl Lyz2Cre/+ (n = 8) BMDMs in a basal state or after 16 hr treatment with 250 μM palmitate. (C) Volcano plot of indicated fatty acid species, expressed as molar percentage of all measured fatty acids, in Pcyt1afl/fl (n = 7) or Pcyt1afl/fl Lyz2Cre/+ (n = 8) BMDMs. (D) Fold induction (compared to BSA alone) of indicated ER stress marker gene and Tnf expression in Pcyt1afl/fl (n = 4) or Pcyt1afl/fl Lyz2Cre/+ (n = 4) BMDMs after 16 hr treatment with 250 μM palmitate, supplemented with or without 25 μM arachidonate.

https://doi.org/10.7554/eLife.47990.022
Figure 6—source data 1

A list of all measured lipid species by LC-MS in Pcyt1afl/fl (n = 7) or Pcyt1afl/fl Lyz2Cre/+ (n = 8) BMDMs in a basal state or after 16 hr treatment with 250 μM palmitate.

Peak areas for each lipid were normalised to the peak areas of a respective internal standard and expressed as peak area ratios.

https://doi.org/10.7554/eLife.47990.027
Figure 6—figure supplement 1
The effects of Pcyt1a deficiency on total PC and PE levels and PE composition in macrophages.

(A) Abundance of all detected PC and PE species in Pcyt1afl/fl (n = 7) or Pcyt1afl/fl Lyz2Cre/+ (n = 8) BMDMs in a basal state. (B) Volcano plot of indicated PE species, expressed as molar percentage of all measured PEs, in Pcyt1afl/fl (n = 7) or Pcyt1afl/fl Lyz2Cre/+ (n = 8) BMDMs in a steady state.

https://doi.org/10.7554/eLife.47990.023
Figure 6—figure supplement 2
Pcyt1a deficiency reduces SREBP1 target gene expression in macrophages.

Normalised expression of SREBP1 target genes in Pcyt1afl/fl (n = 7) or Pcyt1afl/fl Lyz2Cre/+ (n = 8) BMDMs in a basal state.

https://doi.org/10.7554/eLife.47990.024
Figure 6—figure supplement 3
The effects of palmitate treatment on PC to PE ratio and PC composition in macrophages.

(A) PC to PE molar ratio in Pcyt1afl/fl (n = 7) or Pcyt1afl/fl Lyz2Cre/+ (n = 8) BMDMs in a basal state or after 16 hr treatment with 250 μM palmitate. (B) Volcano plot of indicated PC species, expressed as molar percentage of all measured PCs, in Pcyt1afl/fl (n = 7) BMDMs treated with BSA or 250 μM palmitate for 16 hr.

https://doi.org/10.7554/eLife.47990.025
Figure 6—figure supplement 4
Pcyt1a deficiency increases total PUFA levels in macrophages.

Relative abundance of indicated fatty acids in Pcyt1afl/fl (n = 7) or Pcyt1afl/fl Lyz2Cre/+ (n = 8) BMDMs in a basal state or after 16 hr treatment with 250 μM palmitate.

https://doi.org/10.7554/eLife.47990.026
Reduced PC turnover increases membrane PUFA levels in macrophages.

(A3H-arachidonate levels in Pcyt1afl/fl (n = 4) or Pcyt1afl/fl Lyz2Cre/+ (n = 4) BMDMs, pulsed with tracer amounts of 3H-arachidonate for 16 hr and chased with medium for indicated periods of time. (B) Linear regression analysis of the correlation between fold changes in the molar fatty acid percentage induced by genetic Pcyt1a deletion (n = 4) and by PLD inhibition using 15 μM 1-butanol for 24 hr (n = 4).

https://doi.org/10.7554/eLife.47990.028

Tables

Key resources table
Reagent type (species)
or resource
DesignationSource or referenceIdentifiersAdditional
information
Genetic reagent (M. musculus, male)Lepob/obThe Jackson LaboratoryIMSR Cat# JAX:000632, RRID:IMSR_JAX:000632
Genetic reagent (M. musculus)Lyz2Cre/+Clausen et al., 1999IMSR Cat# JAX:004781, RRID:IMSR_JAX:004781Lyz2tm1(cre)Ifo; Donated by Dr. Susan Jackowski
Genetic reagent (M. musculus)Pcyt1afl/flZhang et al., 2000IMSR Cat# JAX:008397, RRID:IMSR_JAX:008397Pcyt1atm1Irt;
Donated by Dr. Susan Jackowski
Antibodyanti-CD45BD Biosciences564279, RRID:AB_2651134Flow cytometry (1:200)
Antibodyanti-CD11bBD Biosciences564443, RRID:AB_2722548Flow cytometry (1:200)
Antibodyanti-SiglecFBD Biosciences562757, RRID:AB_2687994Flow cytometry (1:200)
Antibodyanti-F4/80BioLegend123116, RRID:AB_893481Flow cytometry (1:200)
AntibodyAnti-F4/80 (IHC)Bio-RadCl:A3-1, RRID:AB_1102558IHC(1:100)
Antibodyanti-CD206BioLegend141723, RRID:AB_2562445Flow cytometry (1:200)
Antibodyanti-CD11cBioLegend117336, RRID:AB_2565268Flow cytometry (1:200)
Antibodyanti-Ly6GBioLegend127608, RRID:AB_1186099Flow cytometry (1:200)
Antibodyanti-CD301BioLegend145704, RRID:AB_2561961Flow cytometry (1:200)
Antibodyanti-CCTaAbcamab109263, RRID:AB_10859965WB(1:1000 dilution in 3% milk TBS-T), 4°Covernight
Antibodyanti-beta ActinAbcamab8227, RRID:AB_2305186WB(1:1000 dilution in 3% milk TBS-T), 4°C overnight
Antibodyanti-P-Thr183/Tyr185 JNKCell signalling9251, RRID:AB_331659WB(1:1000 dilution in 3% BSA TBS-T), 4°C overnight
Antibodyanti-JNKCell signalling9252, RRID:AB_2250373WB(1:1000 dilution in 3% BSA TBS-T), 4°C overnight
Antibodyanti-P-Thr180/Tyr182 p38 MAPKCell signalling4511L, RRID:AB_2139679WB(1:1000 dilution in 3% BSA TBS-T), 4°C overnight
Antibodyanti-p38 MAPKCell signalling9212, RRID:AB_330713WB(1:1000 dilution in 3% BSA TBS-T), 4°C overnight
AntibodyAnti-P-Ser473 AKTCell signalling4060, RRID:AB_2315049WB(1:1000 dilution in 3% BSA TBS-T), 4°C overnight
AntibodyAnti-P-Thr308 AKTCell signalling9275, RRID:AB_329828WB(1:1000 dilution in 3% BSA TBS-T), 4°C overnight
AntibodyAnti-pan AKTCell signalling9272, RRID:AB_329827WB(1:1000 dilution in 3% BSA TBS-T), 4°C overnight
AntibodyAnti-GAPDHCell signalling97166, RRID:AB_2756824WB(1:1000 dilution in 3% BSA TBS-T), 4°C overnight
Sequence-based reagentList of nucleotide sequencesThis paperPCR primersSee Table S3 for the full list of qPCR primer sequences
Commercial assay or kitAlpha Trak two glucose meterZoetisN/A
Commercial assay or kitTruSeq Stranded mRNA Library Prep (96 Samples)Illumina20020595
Commercial assay or kitVybrant Phagocytosis Assay KitThermofisherV-6694
Commercial assay or kitRNeasy Mini Kit (250)Qiagen74106
Chemical compound, drugInsulinNovo NordiskEU/1/02/230/003
Chemical compound, drugSTAT 60AMS BiotechCS-502
Chemical compound, drugReverse Transcriptase M-MLVPromegaM170b
Chemical compound, drugM-MLV RT 5x bufferPromegaM351A
Chemical compound, drugMgCl2 25 mMPromegaA351B
Chemical compound, drugRandom PrimersPromegaC118A
Chemical compound, drugdNTP Mix 10 mMPromegaU151B
Chemical compound, drugChlorofomSigma34854
Chemical compound, drugMethanolSigma34860
Chemical compound, drugBF3 MethanolSigmaB1127
Chemical compound, drugHexaneSigma34859
Chemical compound, drugAmmonium formateSigma516961
Chemical compound, drugAcetonitrileVWR83640.320
Chemical compound, drug1-butanolSigma281549
Chemical compound, drugMiltefosineSigmaM5571
Chemical compound, drugArachidonic acidCayman90010
Chemical compound, drugPalmitic acidCayman1000627
Chemical compound, drug[1-14C]- palmitic acidPerkin ElmerNEC075H050UC
Chemical compound, drug[1,2-14C]- acetic acid, sodium saltPerkin ElmerNEC553250UC
Chemical compound, drugMethyl-[3H]- choline chloridePerkin ElmerNET109250UC
Chemical compound, drug[5,6,8,9,11,12,14,15-3H(N)]-arachidonic acidPerkin ElmerNET298Z050UC
Chemical compound, drugBovine Serum AlbuminSigmaA8806BSA used for cell culture experiments
Chemical compound, drugHionic-Fluor scintillation liquidPerkin Elmer6013319
Chemical compound, drugOpti-Fluor scintillation liquidPerkin Elmer6013199
Chemical compound, drugEthanolSigma459836
Chemical compound, drugDAKO Real Peroxidase Blocking solutionAgilentS2023
Chemical compound, drugMOM ImmPress Polymer ReagentVectorMP-2400
Chemical compound, drugDAB Peroxidase (HRP) Substrate KitVectorSK-4100
Chemical compound, drugDako REAL HaematoxylinAgilentS2020
Software, algorithmMetabolizer algorithmCubuk et al., 2018n/ahttp://metabolizer.babelomics.org
Software, algorithmMassHunter Workstation Software Quantitative Analysis (Version B.07.00)Agilent Technologies Incn/a
Software, algorithmThermo Xcalibur Quan browser integration software (Version 3.0)Thermofishern/a
Software, algorithmHALO AIIndica Labsn/a
Software, algorithmTopHat (Version 2.0.11)Kim et al., 2013RRID:SCR_013035
Software, algorithmEdgeRRobinson et al., 2010RRID:SCR_012802
Software, algorithmHiPathiaHidalgo et al., 2017n/ahttp://hipathia.babelomics.org

Data availability

We are submitting raw source data Excel file for LC-MS lipidomics of Pcyt1a-deficient BMDMs, in both palmitate-treated and basal states (containing peak areas for each lipid species normalized to peak areas of respective internal standards) as Figure 6—source data 1. ATM microarray dataset (GSE36669) used in Figure 1 is already published and referenced in this manuscript. We are also submitting a list of differentially expressed genes detected by RNAseq in the eWAT of ob/ob bone marrow transplant mice, with a log (Fold change), log (CPM) and p-value indicated for each gene as Supplementary file 2b. We have uploaded raw RNA sequencing data of liver macrophages isolated from WT and ob/ob mice in the NCBI database under the following accession number: PRJNA541224.

The following data sets were generated
  1. 1
    NCBI BioProject
    1. C Morgantini
    2. M Aouadi
    (2019)
    ID PRJNA541224. Ob/ob liver macrophage RNA sequencing.
The following previously published data sets were used
  1. 1
    NCBI Gene Expression Omnibus
    1. D Montaner
    2. A Vidal-Puig
    3. X Prieur
    4. J Dopazo
    (2012)
    ID GSE36669. Differential lipid partitioning between adipocytes and tissue macrophages modulates macrophage lipotoxicity and M2/M1 polarization in obese mice.

Additional files

Supplementary file 1

The list of biological processes increased in Lepob/ob compared to WT ATMs at week 16 and no change at week 5, ranked in ascending order of adjusted p value.

https://doi.org/10.7554/eLife.47990.029
Supplementary file 2

The list of differentially regulated GO biological processes in eWAT isolated from Lepob/ob BMT Pcyt1afl/fl and Pcyt1afl/flLyz2Cre/+ mice, ranked in ascending order of p value.

Supplementary file 2b. The list of differentially expressed genes in eWAT isolated from Lepob/ob BMT Pcyt1afl/fl and Pcyt1afl/flLyz2Cre/+ mice, ranked in ascending order of p value.

https://doi.org/10.7554/eLife.47990.030
Supplementary file 3

The list of qPCR primer sequences used in this publication.

FAM/TAMRA reporter and quencher detection system was used for genes with indicated probe sequences, and SYBR was used for the remaining genes.

https://doi.org/10.7554/eLife.47990.031
Transparent reporting form
https://doi.org/10.7554/eLife.47990.032

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)

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

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