Infants evaluated in this study were enrolled at birth from the maternity ward at the Site B Clinic in Khayelitsha, Western Cape, South Africa, between 2010 and 2012. This study cohort has been previously described in Ref. (33). Inclusion criteria were term pregnancy (36 weeks gestational age or greater at birth), birth weight >2.4 kg, maternal age 18 years or older, known maternal HIV status, HIV DNA PCR negative at birth or at 6 weeks of age (where applicable), and no potential household tuberculosis contact. All mothers of eligible infants were verbally informed of potential risks and benefits of the study, and signed an IRB-approved consent form in their preferred language, either English or isiXhosa. At birth, 2, 6, 8, and 14 weeks of age, mothers completed an interviewer-assisted survey that included detailed information about the foods consumed by the infant, their health status since the last study visit, including receipt of vaccinations, and maternal health status, including the mother’s most recent CD4 count, if HIV-infected. In addition, infants received a full physical exam, including anthropometrics and evaluation for any oral or skin conditions. Finally, blood (EDTA by venipuncture; BD, CA, USA) was collected at each study visit. Two cohorts of infants were included, one cohort included infants born to HIV-infected mothers and were, therefore, HIV-exposed infants, and a second cohort of HIV-unexposed infants was also included where indicated. All HIV-infected mothers received antiretroviral treatment per South African national guidelines. Demographic data of these two cohorts are presented in Table 1. The low-income setting of this study resulted in difficulties with infant retention, and blood samples at some of the desired time points were not available. There were no demographic differences between infants that were retained in the study and those that were lost to follow-up (33). We acquired the largest number of samples at week 6, and a decreased number of samples at weeks 2 (28 infants), 8 (15 infants), and 14 (11 infants). The majority (>60%) of the infants at weeks 2, 8, and 14 were also evaluated at the week 6 time point.

Infants were defined as breastfed if they had consumed only breastmilk up to the time of sample collection, per maternal report. Non-breastfed infants were infants whose mother’s no longer fed them breast milk (i.e., replacement fed). Infants who consumed both breast milk and non-breast milk foods at the time of sample collection were considered mixed fed.

Blood samples were stored at room temperature in an air conditioned room (temperature no greater than 30°C) for no more than 6 h before transport. Blood was centrifuged, and plasma was aliquoted and frozen at −80°C. Peripheral blood mononuclear cells (PBMCs) were isolated from the remaining cells, using Ficoll (Sigma, MO, USA) density gradient separation. PBMCs were slowly cooled to −80°C in fetal bovine serum (FBS) (Biochrom, UK) + 10% DMSO (Sigma, MO, USA), and transferred to liquid nitrogen for storage.

Frozen PBMCs were thawed at 37°C, washed, and resuspended in PBS (Sigma, MO, USA) + 1% FBS (Biochrom, UK), and cell counts were quantified with a Guava automated cell counter (EMD Millipore, Darmstadt, Germany).

Only infants who had not received BCG vaccination were included in these analyses, the influence of BCG vaccination on immune activation was assessed in a previous study (33). The focus on only non-BCG vaccinated infants enabled us to remove this potential immune stimulation as a confounder from the analysis, as BCG vaccination independently increases T cell activation (33, 34). Each sample was stained with optimized volumes of antibodies to evaluate T cell activation as previously published in Ref. (33). One million cells were stained in a 96-well plate with Live/Dead-Violet (Invitrogen, CA, USA) for 20 min at room temperature, then with anti-CCR5-APC (BD, CA, USA) at 37°C for 15 min, and then stained with anti-CD4-PerCP-Cy5.5 (BD, CA, USA), anti-CD38-PE-Cy7 (eBioscience, CA, USA), anti-HLA-DR-PE (BD, CA, USA), and anti-CD8-QDot605 (Invitrogen, CA, USA) for 20 min at room temperature. Samples were then washed with PBS, permeabilized with CytoFix/CytoPerm (BD, CA, USA), washed twice with Permawash (BD, CA, USA), and then stained intracellularly with anti-Ki-67-FITC and anti-CD3-AlexaFlour700 (BD, CA, USA) for 20 min at room temperature. A final wash was performed and then cells were resuspended in Cell Fix (BD, CA, USA), transferred to a polysterene FACS tube and collected on a Becton Dickinson LSR II flow cytometer within 24 h of staining. All flow cytometry data were compensated and analyzed, using Flow Jo v10 (Treestar, OR, USA). Samples with less than 80% lymphocyte viability were not included in the analysis (Live/Dead negative).

Ochratoxin A was quantified in infant plasma by competitive ELISA in accordance with the manufacturer’s instructions (Helica, CA, USA). Each sample and control was processed in duplicate. Absorbance was measured on a SpectraMax plate reader (Molecular Devices, CA, USA) at 450 nm. OTA levels in each sample were calculated based on a 5-parameter regression of known standards. Values calculated below the lowest standard were assigned the value of the lowest standard. Interassay variability of the assay was minimal (CV 11.2%), based on repeat testing of 10 samples. Lower limit of detection was 80 pg/mL for HIV-exposed infants and 100 pg/mL for HIV-unexposed infants.

Formula and infant food samples were collected from the three non-breastmilk foods most commonly reported by mothers of infants in our study. All samples were advertised as “Baby Cereal,” procured from grocery stores in Cape Town and were manufactured in 2014. Sample 1 was Nestle’s Nestum Baby Cereal Regular and was composed of: wheat flour, brown sugar, flavoring, calcium carbonate, sodium phosphate, vitamins (A, B1, B2, B6, B12, C, D3, E, biotin, niacin, folic acid, pantothenic acid), maltodextrin, sodium chloride, ferrous fumarate, zinc sulfate, Bifidobacterium lactis, and potassium iodide. Sample 2 was Purity’s Baby’s First Cereal Rice and was composed of: rice flour, maltodextrin, non-hydrogenated vegetable fat, tricalcium phosphate, dicalcium phosphate, vitamins (A, C, D, pantothenic acid, thiamine, riboflavin, pyridoxine, biotin, folic acid, cyanocobalamin), and minerals (magnesium, iodine, iron zinc). Sample 3 was Nestle’s Cerelac Baby Cereal with Milk Regular and was composed of: wheat flour, skimmed milk powder, brown sugar, vegetable oil, flavoring, sodium chloride, calcium carbonate, sodium phosphate, maltodextrin, vitamins (A, B1, B2, B6, B12, C, D3, E, biotin, niacin, folic acid, pantothenic acid), ferrous fumarate, B. lactis, zinc sulfate, and potassium iodide. These samples were removed directly from their packaging and tested for OTA by LC-MS/MS, performed by Silliker Inc. Samples were suspended in acetonitrile, ascorbic acid, and water solution and extracted in a stomacher for 2 min. Samples were then cooled for 1 h at 2–8°C and an aliquot was diluted in water, and then vortexed and spun down. The diluted sample was then injected into LC–MS/MS [parent ion (M + H) 404, daughter ions 239 (quantitation trace) and 221 (confirmation trace)]. Lower limit of detection for this assay was 0.8 µg OTA per kilogram of food (LOD 0.8 ppb and LOQ 5 ppb).

Plasma samples from 0-, 2-, and 6-week time points from the infants who were not administered the BCG vaccine were thawed on ice. Prior to cytokine measurements, plasma was centrifuged at 10,000 g for 10 min to remove debris and supernatants were used to determine GM-CSF, IFN-α, IFN-γ, IL-1β, IL-8, CXCL-10 (IP-10), MCP-1, MIP-1β, TNF-α, IL-12p70, and IL-10 concentrations in plasma samples using MilliplexTM Human Cytokine kits (Millipore, MA, USA). The lower limit of detection of these assays range between 3.2 and 7.2 pg/mL for the cytokines measured. Data were collected using a Bio-Plex 200^TM Suspension Array Reader (Bio-Rad Laboratories, CA, USA) and analyzed using Bio-Plex manager software (version 4) (Bio-Rad Laboratories, CA, USA). Cytokine levels that were below the lower limit of detection of the assay were reported as the mid-point between the lowest concentrations measured for each cytokine and 0.

Statistics were performed in Prism 5 (Graphpad, CA, USA) and R (35). Unpaired t-tests were used after testing for normality with D’Agostino and Pearson omnibus normality test (p-value for all populations > 0.05). Where appropriate, multiple comparison correction was performed using the Holme stepdown approach (36). Spearman correlations were used for evaluating correlations. Multivariate analysis was performed in R, using the lm function and the following equation: y=OTA+1.5(gestational age)+male+log10(maternal cd4 count)+log2(visit age), where y = immune activation marker of interest. If the parameter gained normal distribution with transformation, exponential or log transformation was applied to the parameter before further analysis.

To assess plasma levels of the mycotoxin OTA in HIV-exposed, uninfected infants, we evaluated samples from infants enrolled in two studies in Khayelitsha, South Africa (37), an informal settlement just outside of Cape Town, South Africa, which followed infants until 14 weeks of age. A subset analysis, based on availability of samples for these studies, was performed on 2-, 6-, 8-, and 14-week-old infants. There was a slight overrepresentation of female infants in the HIV-exposed infants (Table 1). In the HIV-exposed infants, maternal CD4 count ranged from 71 to over 1,000 cells/mm³, although the median CD4 count was above 200 cells/mm³. Consistent with the eligibility criteria, median gestational age was >38 weeks and birth weights ranged from 2,500 to 4,100 g.

Ochratoxin A plasma levels at 2 weeks of age were relatively low for both breastfed and non-breastfed infants (131 and 151 pg/mL, respectively; Figure 1). However, at 6 weeks of age, OTA plasma levels increased in non-breastfed infants (151 to 250.5 pg/mL), while no increase was observed between 2 and 6 weeks in the exclusively breastfed infants (Figure 1). Longitudinal analysis of non-breast fed infants determined that OTA levels continued to increase in a linear fashion until 8 weeks of age, followed by a slight, but significant, decline at 14 weeks of age, the last time point available for analysis (Figure 2). The OTA levels detected in the plasma of South African infants are within the range of plasma levels observed in adults from OTA-endemic communities (16).

In addition, OTA analysis was also performed on plasma from HIV-unexposed infants at 14 weeks of age. Since many HIV-negative mothers feed their infants both breastmilk and non-breastmilk foods (mixed feeding), we also included these infants in the comparison (Figure 3). Non-breastfed infant plasma contained elevated levels of plasma OTA when compared to both the exclusively breastfed and mixed fed infants (294 and 100 pg/mL, respectively). These levels were comparable to week 14 samples from the HIV-exposed infants (Figure 3, open red squares). Interestingly, mixed fed infants had low to intermediate levels of plasma OTA, suggesting that variable degrees of non-breastmilk food intake results in a broader range of plasma OTA. This finding provides evidence that elevated plasma OTA levels can be observed in both the HIV-exposed and HIV-unexposed infant groups.

To evaluate the immunological effects associated with OTA exposure, the activation status of CD4 T cells was measured in infant PBMCs. Two T cell populations that were of particular interest were HIV target cells (those that co-express the HIV-receptor and co-receptor, CD4 and CCR5) (38) and the activated CD4 T cells (identified by MHC class II molecule HLA-DR expression) (39, 40). CCR5 and HLA-DR expression on CD4 T cells were evaluated at 2, 6, and 8 weeks of age in HIV-exposed, non-breastfed infants. In these infants, OTA plasma levels positively correlated with frequencies of CCR5 (r = 0.44; p = 0.008; Figure 4A) and HLA-DR-expressing CD4 T cells (r = 0.36; p = 0.02; Figure 4B). These findings indicate that increased levels of HIV target cells and CD4+ T cell activation, which promotes HIV replication, is associated with elevated levels of OTA (38, 41). In addition, a multi-analyte profiling panel was utilized to measure plasma levels of nine inflammatory markers. Of these, plasma levels of CXCL10/IP-10 significantly positively correlated with plasma OTA levels (r = 0.58; p = 0.002; Figure 4C). GM-CSF, IFN-α, IL12p70, and IL-10 were undetectable in most samples (LLD = 7.15, 3.2, 3.2, and 3.2 pg/mL, respectively). IFN-γ, IL-8, MCP-1, MIP-1β, TNF-α, and IL-10 levels did not correlate with OTA plasma levels (median = 6.4, 57.3, 556.4, 64.4, and 39.5 pg/mL, respectively). The relationship between OTA levels and CXCL10 persisted after adjusting for multiple comparisons.

To further evaluate these relationships, we performed a stepwise, multivariate regression, including variables that were selected a priori, based on their potential to influence immune activation; maternal HIV disease stage (measured by maternal CD4 count), infant age at sample collection, gender, and gestational age at birth, as well as OTA plasma levels. In this multivariate assessment, these three inflammatory markers (CCR5 and HLA-DR expression on CD4+ T cells and plasma CXCL10 levels) remained significantly associated with plasma levels of OTA (Table 2). Specifically, for every 100 pg/mL increase in plasma OTA levels, the percentage of CD4+ cells expressing CCR5 increased by 1.5-fold, the percentage of CD4+ cells expressing HLA-DR increased by 10-fold, and CXCL10 plasma levels increase by 2.5-fold. Indeed, in contrast to the other factors evaluated, only OTA plasma levels remained significantly correlated with these inflammatory markers.

The findings from this study suggest that South African infants in our study are exposed to OTA via a non-breast milk source. To assess potential sources of OTA in the diet of non-breastfed infants, we measured OTA levels in the four most common, non-breast milk foods given to infants in our study per maternal report. This included the infant formula provided free-of-charge to HIV-infected mothers, and three infant cereals commercially available in Khayelitsha, South Africa. Analysis of OTA levels determined that each of the four infant food samples had less than 0.8 µg OTA per kg of food, the lower limit of detection for the assay. Following this result, we hypothesize that alternative sources of OTA contamination exist.