Human milk antibodies to global pathogens reveal geographic and interindividual variations in IgA and IgG

Abstract

BACKGROUNDThe use of high-throughput technologies has enabled rapid advancement in the knowledge of host immune responses to pathogens. Our objective was to compare the repertoire, protection, and maternal factors associated with human milk antibodies to infectious pathogens in different economic and geographic locations.METHODSUsing multipathogen protein microarrays, 878 milk and 94 paired serum samples collected from 695 women in 5 high and low-to-middle income countries (Bangladesh, Finland, Peru, Pakistan, and the United States) were assessed for specific IgA and IgG antibodies to 1,607 proteins from 30 enteric, respiratory, and bloodborne pathogens.RESULTSThe antibody coverage across enteric and respiratory pathogens was highest in Bangladeshi and Pakistani cohorts and lowest in the U.S. and Finland. While some pathogens induced a dominant IgA response (Campylobacter, Klebsiella, Acinetobacter, Cryptosporidium, and pertussis), others elicited both IgA and IgG antibodies in milk and serum, possibly related to the invasiveness of the infection (Shigella, enteropathogenic E. coli "EPEC", Streptococcus pneumoniae, Staphylococcus aureus, and Group B Streptococcus). Besides the differences between economic regions and decreases in concentrations over time, human milk IgA and IgG antibody concentrations were lower in mothers with high BMI and higher parity, respectively. In Bangladeshi infants, a higher specific IgA concentration in human milk was associated with delayed time to rotavirus infection, implying protective properties of antirotavirus antibodies, whereas a higher IgA antibody concentration was associated with greater incidence of Campylobacter infection.CONCLUSIONThis comprehensive assessment of human milk antibody profiles may be used to guide the development of passive protection strategies against infant morbidity and mortality.FUNDINGBill and Melinda Gates Foundation grant OPP1172222 (to KMJ); Bill and Melinda Gates Foundation grant OPP1066764 funded the MDIG trial (to DER); University of Rochester CTSI and Environmental Health Sciences Center funded the Rochester Lifestyle study (to RJL); and R01 AI043596 funded PROVIDE (to WAP).

Keywords: Adaptive immunity; Immunoglobulins; Infectious disease.

Figure 1. Multipathogen protein microarray principle.

Open reading frame (ORF) expression clone libraries can be constructed from any genome sequence and corresponding source of genomic DNA using high-throughput PCR/recombination cloning. Proteins encoded by the cloned ORF plasmids are expressed using a cell-free in vitro transcription/translation (“IVTT”) system. Each protein is expressed and printed individually onto microarray slides. With as little as 2–5 μL of serum and 100 μL of defatted human milk, complete or partial proteomes can be screened for antibody binding. Isotype-specific bound antibodies are detected using a fluorescently labeled secondary antibody. Using a fluorescence microarray scanner, signal intensities from protein microarrays are acquired and checked for quality, followed by statistical analysis.

Figure 2. Flowchart of samples utilized in the study.

Human milk samples from 6 cohorts in 5 countries were available for assessment of antibody profiles utilizing a protein microarray. As a validation cohort, we sourced an independent set of samples from a mother-infant paired birth cohort from Bangladesh (Cryptosporidium Burden Study) to probe against a mini protein microarray. All the samples available from the smaller studies were assayed at desired time points, and for larger cohorts (200 or more samples), approximately one-third of the samples were selected as representative of the whole cohort or based on the case-control design to include positive and negative infant infectious outcomes where available.

Figure 3. Principal component analysis by economic classification for colostrum and mature human milk.

The scatter plots show principal component (PC) values for each individual’s IgA and IgG responses as points colored by economic region. The top row of plots show responses against enteric pathogen antigens, the second row for respiratory pathogen antigens, and the third row for sepsis-related pathogen antigens. The 2 leftmost columns of plots show colostrum IgA and IgG responses, and the 2 rightmost columns show mature milk IgA and IgG responses. Samples from HICs are shown in orange points, and samples from low- and middle-income countries LMICs are shown in blue points. Colostrum was available from Finland (n = 15) and Pakistan (n = 49); 1 mature milk sample per mother was available from Finland (n = 85), Rochester, New York, U.S. (n = 23), Peru (n = 34), Bangladesh (n = 246), and Pakistan (n = 49). t test P values for PC comparisons between HIC and LMIC are shown in captions below each plot. The 2 PCs with the lowest P values (PC1, PC2 or PC3) for each comparison were plotted.

Figure 4. Pathogen-specific IgA and IgG breadth scores in mature milk and colostrum by economic classification.

The box plots show comparisons of mature milk and colostrum IgA and IgG breadth scores (row headers), defined as the proportion of seropositive (normalized signal ≥ 1.0) antigens per pathogen for each individual (e.g., 20 of 80 positive responses = 0.25 breadth score). The column headers indicate the type of pathogens displayed in each column: enteric, respiratory, and sepsis-related pathogens. Rotavirus and Adenovirus 40/41 were omitted from the enteric pathogens column due to low numbers of reactive antigens. The x axes show each pathogen grouped by HIC (orange boxes) and LMIC (blue boxes) classifications. The y-axes show the IgA or IgG breadth scores on a logarithmic scale with the boxes representing the median and interquartile range. Significant differences by Wilcoxon’s rank sum tests are shown by blue asterisks below each pathogen: *P ≤ 0.05, ** P ≤ 0.005, ** P* ≤ 0.0005. Mature milk samples (n = 438) were included from the latest sample collection for each cohort and were at least 6 weeks postpartum. Colostrum samples (n = 64) were collected in the first 5 days. E. coli, diarrheagenic types EAEC, EPEC and ETEC; RSV, respiratory syncytial virus; GBS, group B Streptococcus.

Figure 5. Comparison of antigen-specific recognition of IgA and IgG for enteric, respiratory, and sepsis pathogens in human milk and serum.

The horizontal bar plots show the number of antigens bound by IgA for each pathogen by sample type and antibody isotype. Pathogens are shown on the y axis, grouped by disease category, with the total number of antigens (“Ag”) present on the multipathogen protein microarray in parentheses. The x axis shows the number of antigens from each pathogen that were reactive. Reactive antigens were defined as antigens with median IgA concentrations of at least 1.0 in normalized signal intensity. Only samples from the Peru and Bangladesh (MDIG) cohorts, which had paired serum and human milk samples at 12 weeks or later postpartum (n = 93 participants; 1 of the 94 participants with paired serum and milk samples did not have a later mature milk sample), were included in this analysis. Note that the total number of antigens differs between the pathogens and therefore the number of reactive antigens does not reflect relative reactivities between the pathogens.

Figure 6. Total IgA and pathogen-specific IgA concentrations decline from colostrum to mature human milk.

(A and B) The line plots show (A) μg/mL of total IgA in human milk and (B) the mean Log₂ signal intensity of IgA antibodies specific for 294 reactive pathogen antigens on the multipathogen protein microarray over 12 to 14 weeks postpartum in Finland (n = 15 subjects), Pakistan (n = 49 subjects), and Peru (n = 9 subjects). The vertical bars represent the SEM. Paired t test P values are shown between time points and colored according to cohort. (C–E) The volcano plots show the difference between pathogen-specific IgA concentrations between time points for (C) Finland and (D and E) Pakistan. Comparison of samples from Peru is not shown due to low number of week 0 colostrum samples (n = 3). Each marker represents an antigen on the multipathogen protein microarray; red open triangles represent IgA responses to individual antigens that are significant after correction for the FDR and black open circles represent IgA responses to individual antigens that were not statistically significant. The x axes show mean differences between time points, and the y axes show the inverse Log₁₀ P value from paired t tests.

Figure 7. Pathogen-specific IgA and IgG concentrations in mature milk and colostrum by economic region.

Scatter plots show IgA or IgG concentrations for each of the reactive antigens from enteric, respiratory, and sepsis-related pathogens (row headers) in mature milk and colostrum (column headers). Antigens classified as “reactive” were those having a median value ≥ 1.0 across the entire study population. Each point represents the mean normalized signal intensity for an individual antigen, colored by pathogen (row legends); solid triangles represent antigens with significant differential reactivity between cohorts by t tests after correction for the FDR. Y axes show means for samples from LMICs, and x axes show means for samples from HICs. The countries included in each plot are listed along the y-axis (LMICs) and x-axis (HICs) with sample sizes in parentheses. The solid diagonal line represents the line of identity (i.e., similar mean signal intensity between LMICs and HICs). Mature milk samples were those from the latest sample collection for each cohort and were at least 6 weeks postpartum. Colostrum samples were collected at 0 weeks. E. coli, diarrheagenic types EAEC, EPEC and ETEC; RSV, respiratory syncytial virus; GBS, group B Streptococcus.

Figure 8. Association of human milk IgA with infection in breastfed infants.

(A) Association of IgA binding to each pathogen for infants subsequently infected with the specific pathogen or not. The log odds from logistic regression of enteric (left) or respiratory (right) infection with increasing IgA binding (x axis) in infants during 1 year and 6 months of follow up, respectively, are shown with the inverse log₁₀ P value (y axis). Associations significant after correction for the FDR are shown in colored triangles. Samples analyzed for diarrheal illness were from the Bangladesh PROVIDE cohort (n = 256) and for respiratory illness were from the Bangladesh MDIG cohort (n = 246). (B) Survival curves of 256 infants from the PROVIDE cohort for enteric pathogens detected by PCR. (C) Hazard ratios of infants during the first year of life divided into the top and bottom halves of mothers’ milk IgA responses for each antigen that was reactive in at least 10% of PROVIDE cohort women. Milk samples included were from mothers with infants that subsequently had pathogen-specific infection. Values below 1.0 represent lower risk of infection in the top half of milk IgA responses compared with the bottom half. For unadjusted P values less than 0.05, antigens were colored (otherwise grey), FDR-adjusted P values less than 0.05 were plotted as triangles. (D and E) Representative Rotavirus A antigen (D) and Campylobacter jejuni antigen (E) corresponding to the samples included in the models shown in the volcano plot (C). The risk tables show the number at risk during 100-day intervals. The Rotavirus A VP4 outer capsid protein is representative of antibodies associated with longer time to infection, while the C. jejuni PEB4 major antigenic peptide (Cj0596) represents antibodies associated with a shorter time to infection. HR, Cox model coefficient for the hazard ratio; CI, confidence interval; P, log-rank test P value; FDR P, adjusted P value.

Figure 9. Validation of human milk IgA correlates of risk in an independent mother-infant birth cohort.

(A) Hazard ratio of infants during the first year of life divided into the top and bottom halves of mothers’ milk IgA responses for antigens included in a validation mini-protein microarray. Milk samples were from women in the independent Cryptosporidium Burden Study cohort (Validation set, n = 144 milk samples). Only proteins that were reactive in at least 10% of women in the Cryptosporidium Burden Study are shown. (B) Hazard ratio and 95% CIs from Cox proportional hazards models of infants during the first year of life divided into the top and bottom half of mothers’ milk IgA responders against Campylobacter jejuni PEB4 and Rotavirus A VP4. P,Cox regression P value; FDR, adjusted P value.