1. Epidemiology and Global Health
Download icon

Enteropathogen antibody dynamics and force of infection among children in low-resource settings

  1. Benjamin F Arnold  Is a corresponding author
  2. Diana L Martin
  3. Jane Juma
  4. Harran Mkocha
  5. John B Ochieng
  6. Gretchen M Cooley
  7. Richard Omore
  8. E Brook Goodhew
  9. Jamae F Morris
  10. Veronica Costantini
  11. Jan Vinjé
  12. Patrick J Lammie
  13. Jeffrey W Priest
  1. University of California, Berkeley, United States
  2. University of California, San Francisco, United States
  3. United States Centers for Disease Control and Prevention, United States
  4. Kenya Medical Research Institute, Centre for Global Health Research, Kenya
  5. Kongwa Trachoma Project, United Republic of Tanzania
  6. Georgia State University, United States
  7. Task Force for Global Health, United States
Research Article
Cite this article as: eLife 2019;8:e45594 doi: 10.7554/eLife.45594
7 figures, 3 tables, 1 data set and 9 additional files

Figures

Figure 1 with 3 supplements
Age-stratified, IgG distributions among a longitudinal cohort of 142 children ages birth to 11 years in Leogane, Haiti, 1990 – 1999.

IgG response measured in multiplex using median fluorescence units minus background (MFI-bg) on the Luminex platform in 771 specimens, marked with rug plots below each distribution. Vertical lines mark seropositivity cutoffs are based on ROC analyses (solid), finite Gaussian mixture models (heavy dash), or distribution among presumed unexposed (light dash). Mixture models failed to converge for ETEC LT B subunit. Created with notebook (https://osf.io/dk54y) and data (https://osf.io/3nv98). Figure 1—figure supplement 1 shows similar distributions from the Tanzania study. Figure 1—figure supplement 2 contrasts Giardia VSP-3 distributions with trachoma pgp3 distributions. Figure 1—figure supplement 3 shows distributions from the Kenyan cohort.

https://doi.org/10.7554/eLife.45594.004
Figure 1—figure supplement 1
Age-stratified, IgG distributions among 4989 children ages 1 to 9 years old in Kongwa, Tanzania, 2012−2015.

IgG response measured in multiplex using median fluorescence units minus background (MFI-bg) on the Luminex platform, marked with rug plots below each distribution. Vertical lines mark seropositivity cutoffs based on ROC analyses (solid) and finite Gaussian mixture models (heavy dash). Mixture models failed to converge for all antibody distributions except for Giardia and E. histolytica. Created with notebook (https://osf.io/dt9zu) and data (https://osf.io/kv4d3).

https://doi.org/10.7554/eLife.45594.005
Figure 1—figure supplement 2
Contrasting age-dependent changes in distributions of IgG levels to Giardia VSP-3 and Chlamydia trachomatis pgp3 antigens among 4989 children ages 1 to 9 years old in Kongwa, Tanzania, 2012 – 2015.

IgG response measured in multiplex using median fluorescence units minus background (MFI-bg) on the Luminex platform, marked with rug plots below each distribution. Vertical lines mark seropositivity cutoffs based on ROC analyses (solid) and finite Gaussian mixture models (dash). Created with notebook (https://osf.io/dt9zu) and data (https://osf.io/kv4d3).

https://doi.org/10.7554/eLife.45594.006
Figure 1—figure supplement 3
IgG distributions among children ages 4 to 17 months old in Asembo, Kenya, 2013.

IgG response measured in multiplex using median fluorescence units minus background (MFI-bg) on the Luminex platform. N = 439 samples from 240 children, marked with rug plots below each distribution. Vertical lines mark seropositivity cutoffs based on ROC analyses (solid), finite Gaussian mixture models (heavy dash), or distribution among presumed unexposed (light dash). Mixture models failed to converge for ETEC LT B subunit and Campylobacter p18. Created with notebook (https://osf.io/456jp) and data (https://osf.io/2q7zg).

https://doi.org/10.7554/eLife.45594.007
Joint distributions of select enteric pathogen antibody responses among children in three cohorts from Haiti, Kenya, and Tanzania.

Each panel includes Spearman rank correlations (ρ) and locally weighted regression smoothers with default parameters, trimmed to 95% of the data to avoid edge effects. Antibody response measured in multiplex using median fluorescence units minus background (MFI-bg) on the Luminex platform. Empty panels indicate that the antibodies were not measured in that cohort. Supplementary file 3 includes all pairwise comparisons. Created with notebook (https://osf.io/hv9ce) and data (https://osf.io/3nv98, https://osf.io/2q7zg, https://osf.io/kv4d3).

https://doi.org/10.7554/eLife.45594.008
Figure 3 with 2 supplements
Age dependent mean IgG and seroprevalence in Haiti.

Geometric means (A) and seroprevalence (B), estimated with cubic splines among children ages birth to 10 years in Leogane, Haiti 1990–1999. Shaded bands are approximate, simultaneous 95% confidence intervals. IgG response measured in multiplex using median fluorescence units minus background (MFI-bg) on the Luminex platform (N = 771 measurements from 142 children). Created with notebook (https://osf.io/jeby3) and data (https://osf.io/3nv98). Data for some antigens measured among children < 5 years previously published (Arnold et al., 2017). Figure 3—figure supplement 1 includes similar curves from Kenya; Figure 3—figure supplement 2 includes similar curves from Tanzania.

https://doi.org/10.7554/eLife.45594.009
Figure 3—figure supplement 1
Age dependent mean IgG and seroprevalence in Kenya.

Geometric means (A) and seroprevalence (B), estimated with cubic splines among children ages 4 to 17 months in Asembo, Kenya, 2013. Shaded bands are approximate, simultaneous 95% confidence intervals. Ages were measured in months completed (rounded) so points in the figure are jittered. IgG response measured in multiplex using median fluorescence units minus background (MFI-bg) on the Luminex platform (N = 439 measurements from 240 children). Created with notebook (https://osf.io/jeby3) and data (https://osf.io/2q7zg).

https://doi.org/10.7554/eLife.45594.010
Figure 3—figure supplement 2
Age dependent mean IgG and seroprevalence in Tanzania.

Geometric means (A) and seroprevalence (B), estimated with cubic splines among children ages 1 to 9 years in Kongwa, Tanzania, 2012–2015. Shaded bands are approximate, simultaneous 95% confidence intervals. Ages were measured in years completed (rounded) so points in the figure are jittered. IgG response measured in multiplex using median fluorescence units minus background (MFI-bg) on the Luminex platform in 4989 specimens. Salmonella and Campylobacter antigens were only included in the multiplex in 2012 (N = 902 specimens). Seropositivity cutoffs could not be estimated for bacterial pathogens in this study, so seroprevalence curves are not shown. Created with notebook (https://osf.io/jeby3) and data (https://osf.io/kv4d3).

https://doi.org/10.7554/eLife.45594.011
Figure 4 with 3 supplements
Longitudinal changes in IgG response over six repeated measurements among 142 children ages 0–11 years in Leogane, Haiti, 1990 – 1999.

Measurements were spaced by approximately 1 year (median spacing = 1, IQR = 0.7, 1.3). Horizontal dashed lines mark seropositivity cutoffs for each antibody. The number of children measured at each visit was: n1 = 142, n2 = 142, n3 = 140, n4 = 131, n5 = 111, n6 = 66); 29 children had >6 measurements that are not shown. IgG response measured in multiplex using median fluorescence units minus background (MFI-bg) on the Luminex platform. Created with notebook (https://osf.io/vyhra), which includes additional visualizations, and data (https://osf.io/3nv98). Figure 4—figure supplement 1 includes a similar figure from the Kenya cohort. Figure 4—figure supplements 2 and 3 summarize the proportion of children in each category across measurement rounds in Haiti and Kenya.

https://doi.org/10.7554/eLife.45594.012
Figure 4—figure supplement 1
Longitudinal changes in IgG response between enrollment and follow-up among 205 children ages 4–17 months in Asembo, Kenya, 2013.

Horizontal dashed lines mark seropositivity cutoffs for each antibody. IgG response measured in multiplex using median fluorescence units minus background (MFI-bg) on the Luminex platform. Created with notebook (https://osf.io/6gk2q), which includes additional visualizations, and data (https://osf.io/2q7zg).

https://doi.org/10.7554/eLife.45594.013
Figure 4—figure supplement 2
Proportion of children ages 0–11 years with different longitudinal changes in IgG response over six repeated measurements in Leogane, Haiti, 1990 – 1999.

Measurements were spaced by approximately 1 year (median spacing = 1, IQR = 0.7, 1.3). The number of children measured at each visit was: n1 = 142, n2 = 142, n3 = 140, n4 = 131, n5 = 111, n6 = 66); 29 children had >6 measurements that are not shown. IgG response measured in multiplex using median fluorescence units minus background (MFI-bg) on the Luminex platform. Created with notebook (https://osf.io/vyhra) and data (https://osf.io/3nv98).

https://doi.org/10.7554/eLife.45594.014
Figure 4—figure supplement 3
Proportion of children ages 4–17 months with different longitudinal changes in IgG response between enrollment and follow-up 6 months later in Asembo, Kenya, 2013.

N = 205 children measured longitudinally. IgG response measured in multiplex using median fluorescence units minus background (MFI-bg) on the Luminex platform. Created with notebook (https://osf.io/6gk2q) and data (https://osf.io/2q7zg).

https://doi.org/10.7554/eLife.45594.015
Figure 5 with 1 supplement
Average force of infection versus seroprevalence for enteropathogens measured in the Kenya and Haiti cohorts.

Force of infection estimated from prospective seroconversion rates. Vertical lines indicate 95% confidence intervals. Created with notebook (https://osf.io/jp9kf) and data (https://osf.io/2q7zg, https://osf.io/3nv98). Figure 5—figure supplement 1 includes estimates from Haiti stratified by age bands.

https://doi.org/10.7554/eLife.45594.016
Figure 5—figure supplement 1
Average force of infection versus seroprevalence for enteropathogens measured in the Haiti cohort, stratified by different age bands.

Force of infection estimated from prospective seroconversion rates. Vertical lines indicate 95% confidence intervals. Created with notebook (https://osf.io/jp9kf) and data (https://osf.io/3nv98).

https://doi.org/10.7554/eLife.45594.017
Enteropathogen seroconversion and seroreversion rates among 205 children ages 4 to 17 months measured longitudinally in Asembo, Kenya, 2013.

The seroconversion rate is a measure of a pathogen's force of infection. Longitudinal estimates are non-parametric rates of incident seroconversions and seroreversions among children at risk, assumed to occur at the midpoint of the measurement interval. Cross-sectional estimators were derived from age-specific seroprevalence curves using semiparametric cubic splines (spline), a reversible catalytic model (RCM) that assumed constant seroconversion and seroreversion rates with the seroreversion rate estimated from prospective data, and a parametric constant rate survival model (exponential). Error bars mark 95% confidence intervals. IgG response measured in multiplex using median fluorescence units minus background (MFI-bg) on the Luminex platform (N = 410 measurements from 205 children). Created with notebooks (https://osf.io/sqvj7, https://osf.io/j9nh3) and data (https://osf.io/2q7zg).

https://doi.org/10.7554/eLife.45594.019
Empirical seroconversion rates compared with estimates from 100 simulated datasets with daily resolution IgG trajectories that were sampled at a 30 day interval before estimating seroconversion rates.

Vertical lines in the simulation results indicate medians. (A) In the Kenya cohort, children were ages 4–18 months and empirical IgG measurements were measured every 6 months (approximately 180 days). (B) In the Haiti cohort, children were ages 0–11 years, and empirical IgG measurements were measured approximately each year (median spacing = 1 year, IQR = 0.7, 1.3). Created with notebooks (https://osf.io/qmdf2, https://osf.io/9fxhb) and data (https://osf.io/2q7zg, https://osf.io/3nv98).

https://doi.org/10.7554/eLife.45594.020

Tables

Table 1
Number of children and samples tested, and estimated seropositivity cutoffs by country and antigen included in the seroepidemiologic analyses.
https://doi.org/10.7554/eLife.45594.003
Seropositivity cutoff, log10 IgG (MFI-bg) *
NNExternalMixturePresumed
childrensamplesReferenceModelUnexposed
Leogane, Haiti
Giardia VSP-31427712.421.642.11
Giardia VSP-51427712.311.461.88
Cryptosporidium Cp171427712.262.002.58
Cryptosporidium Cp231427712.702.752.57
E. histolytica LecA1427712.482.301.93
Salmonella LPS group B1427711.601.37
Salmonella LPS group D1427711.482.48
ETEC LT B subunit1427712.86
Norovirus GI.41427712.512.09
Norovirus GII.4.NO1427712.042.24
Asembo, Kenya
Giardia VSP-32404452.811.621.67
Giardia VSP-52404452.651.821.67
Cryptosporidium Cp172404452.632.582.38
Cryptosporidium Cp232404453.143.402.36
E. histolytica LecA2404451.89
Salmonella LPS group B2404451.36
Salmonella LPS group D2404451.41
ETEC LT B subunit2404452.79
Cholera toxin B subunit2404452.91
Campylobacter p182404452.11
Campylobacter p392404452.612.57
Kongwa, Tanzania
Giardia VSP-3498949892.232.04
Giardia VSP-5498949892.152.27
Cryptosporidium Cp17498949892.26
Cryptosporidium Cp23498949892.58
E. histolytica LecA498949891.972.50
Salmonella LPS group B902902
Salmonella LPS group D902902
ETEC LT B subunit49894989
Cholera toxin B subunit40874087
Campylobacter p18902902
Campylobacter p39902902
  1. *Seropositivity cutoffs determined using external reference samples (typically ROC curves except for Giardia and E. hystolitica in Haiti), finite Gaussian mixture models, or distribution among the presumed unexposed (see Materials and methods for details). External reference cutoffs vary across cohorts for the same antigen due to use of different bead sets in each cohort. External reference cutoffs reported from years (2013–2015) in Tanzania, estimated among 4087 samples. Cutoff values are missing if they could not be estimated in each method; cutoff values based on the presumed unexposed required longitudinal measurements within individual children and therefore could not be estimated for any antigen in the repeated cross-sectional design in Tanzania.

    Measured only in year 1 of the study (2012).

  2. Measured only in years 2–4 of the study (2013–2015).

Table 2
Incidence rates of seroconversion and seroreversion per child year among children ages 0–11 years in Haiti, 1990–1999.
https://doi.org/10.7554/eLife.45594.018
Seropositivity cutoff *4-Fold change in IgG levels †
PathogenChild- yearsIncident casesRate
(95% CI)
Child- yearsIncident casesRate
(95% CI)
Ratio of casesRatio of rates
Seroconversion/boosting
Giardia VSP-3 or VSP-5269.61080.40 (0.34, 0.48)277.21200.43 (0.35, 0.54)1.11.1
Cryptosporidium Cp17 or Cp23109.3700.64 (0.54, 0.77)241.02040.85 (0.73, 0.97)2.91.3
E. histolytica LecA283.7970.34 (0.28, 0.42)297.11070.36 (0.29, 0.45)1.11.1
Salmonella LPS groups B or D132.1750.57 (0.47, 0.68)226.81490.66 (0.54, 0.80)2.01.2
 ETEC LT B subunit9.7111.13 (0.75, 1.82)32.1321.00 (0.70, 1.45)2.90.9
 Norovirus GI.4213.0800.38 (0.30, 0.47)254.31070.42 (0.34, 0.53)1.31.1
 Norovirus GII.4.NO105.8670.63 (0.51, 0.80)147.21000.68 (0.54, 0.86)1.51.1
Seroreversion/waning
Giardia VSP-3 or VSP-5441.6910.21 (0.17, 0.25)290.91270.44 (0.35, 0.54)1.42.1
Cryptosporidium Cp17 or Cp23586.1290.05 (0.03, 0.07)273.51710.63 (0.53, 0.74)5.912.6
E. histolytica LecA395.2430.11 (0.08, 0.15)310.4670.22 (0.16, 0.27)1.62.0
Salmonella LPS groups B or D544.3250.05 (0.03, 0.07)344.5890.26 (0.20, 0.32)3.65.6
 ETEC LT B subunit702.120.00 (0.00, 0.01)649.7220.03 (0.02, 0.05)11.011.9
 Norovirus GI.4464.9280.06 (0.03, 0.09)362.2560.15 (0.11, 0.21)2.02.6
 Norovirus GII.4.NO574.3190.03 (0.02, 0.05)477.9390.08 (0.06, 0.11)2.12.5
  1. *Incident changes in serostatus defined by crossing seropositivity cutoffs.

    Incident changes in serostatus defined by a 4-fold increase or decrease in IgG levels (MFI-bg), with incident boosting episodes restricted to changes that ended above the seropositivity cutoff and incident waning episodes restricted to changes that started from above the seropositivity cutoff.

Key resources table
Reagent type
or resource
DesignationSource or referenceIdentifiersAdditional
information
Peptide, recombinant proteinGiardia intestinalis VSP-3PMID: 17901334
PMID: 20876825
GenBank: XM_001707314Dr. Jeffrey Priest (CDC)
Peptide, recombinant proteinGiardia intestinalis VSP-5PMID: 11500396
PMID: 20876825
GenBank: AF354538.1Dr. Jeffrey Priest (CDC)
Peptide, recombinant proteinCryptosporidium Cp17PMID: 10699255
PMID: 15165066
GenBank: AF114166Dr. Jeffrey Priest (CDC)
Peptide, recombinant proteinCryptosporidium Cp23PMID: 8892291
PMID: 10203492
GenBank: U34390Dr. Jeffrey Priest (CDC)
Peptide, recombinant proteinEntamoeba histolytica LecAPMID: 2000392
PMID: 14741152
PMID: 24591430
GenBank: M60498Dr. William Petri (University of Virginia) and Dr. Joel Herbein (TechLab)
Peptide, recombinant proteinCampylobacter jejuni p18PMID: 8576327
PMID: 16014430
GenBank: X83374Dr. Jeffrey Priest (CDC)
Peptide, recombinant proteinCampylobacter jejuni p39PMID: 10688204
PMID: 16014430
GenBank: CAL34198.1Dr. Jeffrey Priest (CDC)
Peptide, recombinant proteinETEC heat labile toxin B subunitPMID: 3882744
PMID: 18494692
Sigma-Aldrich
Peptide, recombinant proteinCholera toxin B subunitPMID: 3882744Sigma-Aldrich
Peptide, recombinant proteinSalmonella enterica LPS group BPMID: 2567429
PMID: 17329442
Sigma-Aldrich
Peptide, recombinant proteinSalmonella enterica LPS group DPMID: 2567429
PMID: 17329442
Sigma-Aldrich
Peptide, recombinant proteinNorovirus GI.4
Virus Like Particles
This paperDr. Jan Vinje (CDC)
Peptide, recombinant proteinNorovirus GII.4.NO
Virus Like Particles
This paperDr. Jan Vinje (CDC)

Data availability

Analyses were conducted in R version 3.5.3. Data and computational notebooks used to complete the analyses are available through GitHub (https://github.com/ben-arnold/enterics-seroepi; copy archived at https://github.com/elifesciences-publications/enterics-seroepi) and the Open Science Framework (osf.io/r4av7).

The following data sets were generated
  1. 1
    The Open Science Framework
    1. BF Arnold
    2. DL Martin
    3. J Juma
    4. H Mkocha
    5. JB Ochieng
    6. GM Cooley
    7. Omore R Richard
    8. EB Goodhew
    9. JF Morris
    10. V Costantini
    11. J Vinjé
    12. PJ Lammie
    13. JW Priest
    (2019)
    Data and computational notebooks used to complete the analyses in Enteropathogen antibody dynamics and force of infection among children in low-resource settings.
    https://doi.org/10.17605/osf.io/r4av7

Additional files

Supplementary file 1

Effect of intervention, bead lot, and season on enteropathogen antibody response (osf.io/6br2f).

https://doi.org/10.7554/eLife.45594.021
Supplementary file 2

Classification agreement between different seropositivity cutoff approaches (osf.io/7x6sw).

https://doi.org/10.7554/eLife.45594.022
Supplementary file 3

Joint distributions of antibody response (osf.io/wchzq).

https://doi.org/10.7554/eLife.45594.023
Supplementary file 4

IgG measurements in the Kenya cohort among children with- and without confirmed Cryptosporidium and Giardia infections in diarrheal stools (osf.io/e4tbg).

https://doi.org/10.7554/eLife.45594.024
Supplementary file 5

Sensitivity analyses: fold-changes in IgG used to identify presumed unexposed measurements and force of infection in Haiti and Kenya (osf.io/u79bm).

https://doi.org/10.7554/eLife.45594.025
Supplementary file 6

Estimation of age-dependent means and seroprevalence using multiple approaches (osf.io/r25hp).

https://doi.org/10.7554/eLife.45594.026
Supplementary file 7

Estimation of force of infection from age-structured seroprevalence in Kenya (osf.io/9wbh5).

https://doi.org/10.7554/eLife.45594.027
Supplementary file 8

Simulation study to assess the influence of sampling intervals on serological estimates of force of infection (osf.io/9zt4d).

https://doi.org/10.7554/eLife.45594.028
Transparent reporting form
https://doi.org/10.7554/eLife.45594.029

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)