Early-Life Iron Deficiency and Its Natural Resolution Are Associated with Altered Serum Metabolomic Profiles in Infant Rhesus Monkeys

ABSTRACT

Background

Iron deficiency is the most common nutrient deficiency in human infants aged 6 to 24 mo, and negatively affects many cellular metabolic processes, including energy production, electron transport, and oxidative degradation of toxins. There can be persistent influences on long-term metabolic health beyond its acute effects.

Objectives

The objective was to determine how iron deficiency in infancy alters the serum metabolomic profile and to test whether these effects persist after the resolution of iron deficiency in a nonhuman primate model of spontaneous iron deficiency.

Methods

Blood was collected from naturally iron-sufficient (IS; n = 10) and iron-deficient (ID; n = 10) male and female infant rhesus monkeys (Macaca mulatta) at 6 mo of age. Iron deficiency resolved without intervention upon feeding of solid foods, and iron status was re-evaluated at 12 mo of age from the IS and formerly ID monkeys using hematological and other indices; sera were metabolically profiled using HPLC/MS and GC/MS with isobaric standards for identification and quantification at both time points.

Results

A total of 413 metabolites were measured, with differences in 40 metabolites identified between IS and ID monkeys at 6 mo (P 0.05). At 12 mo, iron-related hematological parameters had returned to normal, but the formerly ID infants remained metabolically distinct from the age-matched IS infants, with 48 metabolites differentially expressed between the groups. Metabolomic profiling indicated altered liver metabolites, differential fatty acid production, increased serum uridine release, and atypical bile acid production in the ID monkeys.

Conclusions

Pathway analyses of serum metabolites provided evidence of a hypometabolic state, altered liver function, differential essential fatty acid production, irregular uracil metabolism, and atypical bile acid production in ID infants. Many metabolites remained altered after the resolution of ID, suggesting long-term effects on metabolic health.

Introduction

Iron deficiency with or without anemia is common in human infants aged 6 to 24 mo, and is a risk factor for long-term behavioral and cognitive problems (1). Iron is essential for the synthesis of neurotransmitters, synaptogenesis, and myelination in the brain (1, 2). Iron and iron-containing enzymes are also critical for many cellular metabolic processes, including energy production, electron transport, and oxidative degradation of toxins, and serve as required cofactors for tryptophan pyrrolase and phosphoenolpyruvate carboxykinase activity (3).

Total body iron is heavily prioritized to hemoglobin (Hgb) synthesis in RBCs, with only ∼30% of the iron allocated to other tissues (4). During negative iron balance in infants, microcytic anemia becomes the final clinical manifestation of iron deficiency, occurring after available iron has already been prioritized to the RBCs over all other organs, including the brain (5, 6). This interorgan and tissue prioritization is most evident in the late fetal and early infancy periods and follows a specific sequence, with the liver, kidneys, and skeletal muscle becoming iron depleted prior to the heart and brain (7–10). A similar pattern of prioritization occurs during recovery from anemia, with the heme compartment prioritized over replenishment of tissue iron in other organs (9).

Metabolomic analysis provides a comprehensive physiological snapshot suitable for simultaneously interrogating multiple metabolic processes in a cell or organ and thus can be used to discern aberrant metabolic consequences of nutrient deficiencies and other insults (11–14). We have previously demonstrated that iron deficiency is associated with an altered metabolomic profile of the cerebrospinal fluid (CSF) compartment in infant monkeys and found that these metabolomic alterations preceded the onset of anemia and persisted for months after its natural resolution (9, 15). However, CSF is a difficult biofluid to obtain in clinical practice. In contrast, blood collection is routine, and serum markers can provide insight into both systemic metabolic processes and the co-occurring metabolic alterations in the brain. Whereas the CSF metabolomic profile reflects the state of the central nervous system (15), the more easily collected serum metabolomic profile can provide a comprehensive picture of the multiple organs affected by iron deficiency (16). Plasma and serum metabolomic analyses have proven useful for diagnosing many clinical conditions and monitoring their progression or response to treatment (14, 17–19). A recent population-based study demonstrated that iron deficiency is associated with altered plasma metabolomic profiles in adult humans, with fatty acid species, cholesterol, branched-chain amino acid catabolites, and catabolites of heme being particularly affected (20). However, the higher metabolic state of maturing organs during the rapid growth phase of infancy, especially in the context of insufficient iron resources, differs greatly from the adult (21). It is our hypothesis that the requirements for nutrients, such as iron, that support oxidative metabolism in all organs, as well as the effect of iron deficiency on those organs would be greater in an infant (13).

To gain further insight into these metabolic effects, we used a well-characterized monkey model of spontaneous infantile iron deficiency that mirrors the clinical presentation and time course in human infants. Rhesus monkeys are an optimal model because of their genetic relatedness, and comparable iron requirements during the nursing period (21–26). Human and rhesus infants also exhibit identical metabolites in serum (22), indicating that evidence for an iron deficiency–induced altered metabolic function in this monkey model will have translational relevance to human infants.

We employed a targeted analysis with a straightforward experimental design and directed hypotheses. The primary comparison was a pairwise contrast of sera from iron-deficient (ID) infant monkeys compared with age-matched, iron-sufficient (IS) infant monkeys. Iron deficiency was not specifically treated, but allowed to resolve naturally after weaning with the feeding of iron-containing solid foods. Only 1 secondary comparison was conducted, which was to determine if the metabolic state of the formerly ID monkeys still differed from the IS monkeys when they were 1 y old. Our findings indicate that important metabolic pathways are affected in ID infants and that many of the pathways remained altered after recovery of iron deficiency at 1 y of age, which is the age at which the iron status of human infants is first evaluated during a well-baby clinic visit (27).

Methods

Subjects

This prospective cohort study identified 20 healthy infant rhesus monkeys (Macaca mulatta), 10 IS subjects (4 males and 6 females) and 10 ID subjects (7 males and 3 females), at 6 mo of age. Infant rhesus monkeys were scheduled for blood sampling at 6 and 12 mo of age and categorized by iron status, but otherwise selected at random to be representative of other monkeys born in a large indoor primate colony at University of Wisconsin—Madison. All were singleton infants born to 20 different mothers after a full-term pregnancy and vaginal delivery. All were mother-reared under standardized husbandry and diet conditions as previously described (9, 13, 28). The experimental protocol and specimen collection were approved by the institutional Animal Care and Use Committee at the University of Wisconsin—Madison.

Housing and diet

Mothers were fed a commercial biscuit diet (5LFD; LabDiet) (29) and supplemented 3–4 times per week with fruits and vegetables. The composition of the diet has been published previously (12, 29) and relevant nutrient concentrations are provided in Supplemental Table 1. The iron concentration in this diet (225 mg/kg diet) is adequate for a healthy nonpregnant adult monkey, but not sufficient to entirely meet the additional maternal iron needs during pregnancy. Although the biscuits are derived primarily from plant sources, fish meal is an ingredient indicating that the diet might contain a mixture of heme and nonheme iron. As described previously, ∼20–30% of infants born to mothers consuming this diet will develop iron deficiency by 6 mo due to the combined influence of lower storage iron at birth and insufficient iron in breast milk to entirely meet the higher iron needs during postnatal growth (9). During this period of iron deficiency, prior validations of the model have determined that serum iron was decreased to 15 µmol/L compared with a normal concentration of 20 µmol/L; transferrin saturation (TSAT) was significantly below normal at 18% compared with 32%; and ferritin concentrations were only 10 pmol/L compared with 25 pmol/L in an IS infant (28, 29). Infants were nursed until the first blood sample at 6 mo of age, after which they were weaned into identical settings. They were relocated into small social groups of 4–8 weanling infant monkeys and fed solid food through 12 mo of age. The same 5LFD biscuits were provided to these older infants as the primary diet.

Specimen collection

Blood samples (<3 mL) were collected between 09:30 and 11:00 from the infants in their home environment at 6 and 12 mo of age via femoral venipuncture and analyzed as single individuals, not pooled. The blood was used to determine hematological parameters [Hgb, RBC count, mean corpuscular volume (MCV), hematocrit (HCT), mean corpuscular hemoglobin (MCH), mean corpuscular hemoglobin concentration (MCHC), red blood cell distribution width (RDW), and the iron-specific index of zinc protoporphyrin/heme (ZnPP/H)] (28, 29). These erythrocyte iron indices were previously found to be strongly associated with serum ferritin and TSAT in infant monkeys and have the advantages of requiring minimal blood volume and being less dependent upon collection time of day (28, 29). The remaining blood was centrifuged at 2,061 x g at room temperature for 12 min to generate serum, which was frozen in an ultracold freezer at < −70°C until the metabolomic analysis.

Hematological assessment

Hgb, RBC count, MCV, HCT, MCH, MCHC, and RDW were determined at a CLIA (Clinical Laboratory Improvement Amendments) certified clinical laboratory with prior experience in processing samples from nonhuman primates (Meriter Labs). ZnPP/H, reflecting incomplete iron incorporation into the erythrocyte, was determined on site using a hematofluorometer (Aviv Biomedical). Criteria for determining iron deficiency at both time points were ≥3 of the following: Hgb <110 g/L, MCV <60 fL, ZnPP/H >150 µmol/mol, and RDW >15%, as defined in previous studies (9, 13, 28–30).

Metabolomic analysis

Sample preparation was conducted using a proprietary series of organic and aqueous extractions to remove the protein fraction while allowing maximum recovery of small molecules (Precision Metabolomics). The resulting extract was divided into 2 fractions; 1 for analysis by HPLC and 1 for analysis by GC. The HPLC/MS portion of the platform was based on a Waters Acquity UPLC and a Thermo-Finnigan LTQ mass spectrometer, which consisted of an electrospray ionization source and linear ion-trap mass analyzer. The sample extract was split into 2 aliquots, dried, then reconstituted in acidic or basic HPLC-compatible solvents, each of which contained 11 or more injection standards at fixed concentrations. One aliquot was analyzed using acidic positive ion–optimized conditions and the other using basic negative ion–optimized conditions in 2 independent injections using separate dedicated columns. The MS analysis alternated between MS and data-dependent tandem mass spectrometry (MS²) scans using dynamic exclusion.

The samples destined for GC/MS analysis were redried under vacuum desiccation for a minimum of 24 h prior to being derivatized under dried nitrogen using bistrimethyl-silyl-trifluoroacetamide. The GC column was 5% phenyl and the temperature ramp from 40°C to 300°C in a 16-min period. Samples were analyzed on a Thermo-Finnigan Trace DSQ fast-scanning single-quadrupole mass spectrometer using electron impact ionization (31).

Statistical analysis

Penalized canonical correlation analysis (CCA) was performed using the PMA package for the R statistical computing environment, to assess the overall association of the metabolomic profile with iron deficiency with a single statistical test (32). CCA is a technique used to assess the relation between 2 sets of measurements, by identifying a weighted sum in 1 set (here, metabolites) that correlates with a weighted sum in the other set (here, hematological parameters). We assessed the significance of this correlation via leave-one-out cross-validation. Random effects ANOVA analysis was used to assess specific contrasts for each metabolite: IS compared with ID at 6 mo, IS compared with ID at 12 mo, IS at 6 mo compared with IS at 12 mo, and ID at 6 mo compared with IS at 12 mo. To reduce the number of statistical tests, the meaningfulness of observed differences in metabolites was ascertained by considering the individual analytes within pathway models, applying Ingenuity Pathway Analysis (IPA; Qiagen) to the resulting data as previously described (33, 34). IPA predicts pathway models using the right-tailed Fisher exact test as previously described (33, 34). The data extraction of the raw MS data files yielded information that could be loaded into a relational database and manipulated. Once in the database, the information was examined, and appropriate quality control limits were imposed. Peaks were identified using proprietary peak integration software, and component parts were stored in a separate and specifically designed complex data structure (Metabolon). Compounds were identified by comparison with library entries of purified standards. The combination of chromatographic properties and mass spectra gave an indication of a match to the specific compound as previously described (31). No sample size determination was performed a priori, although based on prior findings and effect sizes, the planned comparison was on 10 subjects per condition. The analysis included the metabolite profile of the IS infants at 6 mo of age as the reference point. Thus, the primary research question was addressed with a single targeted analysis to determine differences between the ID and IS infants. The second model compared the nature of metabolite changes in the same infants 6 mo later, and also a directed analysis of differences with respect to the IS infants.

Results

Ten otherwise healthy infants were identified as being ID at 6 mo of age based on ≥3 of the 4 following indices: Hgb <110 g/L, MCV <60 fL, ZnPP/H >150 µmol/mol, and RDW >15%, compared with IS monkeys at 6 mo of age (Figure 1A, B, C, and D, respectively). The 10 IS infants did not experience iron deficiency throughout the first year of life based on Hgb concentration, MCV, ZnPP/H, and RDW values.

FIGURE 1.

Serum Hgb concentration (A), MCV (B), ZnPP/H (C), and RDW (D) in rhesus monkeys that were IS or ID at 6 mo of age prior to being fed solid food, and at 12 mo of age. Values are mean SEM, n = 10. Main effects of age, iron status, and age × iron status, P < 0.05, 2-factor ANOVA. Groups without a common letter differ, P < 0.05. Hgb, hemoglobin; ID, iron-deficient; IS, iron-sufficient; MCV, mean corpuscular volume; RDW, red blood cell distribution width; ZnPP/H, zinc protoporphyrin/heme.

At 6 mo of age, CCA identified a weighted sum of the hematology parameters that was correlated with a weighted sum of the metabolites. The significance and robustness of the CCA results were assessed via leave-one-out cross-validation. Under leave-one-out cross-validation, the data for 1 monkey were removed (i.e., the test case), the weights for hematology and metabolites were estimated from the remaining monkeys (i.e., the training cases), and the resulting sums were computed for the test case to generate its hematology and metabolite scores. This process was repeated with each monkey as the test case. This cross-validation confirmed the significant correlation between the weighted hematology and metabolite scores (r = 0.67; P < 0.001). Moreover, there was a significant separation of the IS and ID monkeys for both the hematology scores (t = −8.83; P < 0.001) and metabolite scores (t = −3.61; P = 0.002) (Figure 2A). By 12 mo of age, iron deficiency had resolved, and both groups of monkeys were now more similar from a hematological perspective (t = −1.90; P = 0.08). Nevertheless, the metabolite profiles remained distinct (t = −3.13; P = 0.006) (Figure 2B). The histogram of P values across metabolites is shown for each group-wise comparison (Supplemental Figure 1). Under a null scenario in which no metabolites are associated, the distribution of P values will be approximately uniform, indicated by the dashed line (Supplemental Figure 1). Comparisons related to the resolution of iron deficiency (ID 6 mo compared with 12 mo) and iron status (ID 6 mo compared with IS 6 mo, and ID 12 mo compared with IS 12 mo) demonstrated a higher occurrence of significantly different metabolites with low P values than was evident after normal development alone (IS at 6 mo and 12 mo) (Supplemental Figure 1).

FIGURE 2.

This canonical correlation analysis weighted distribution demonstrating a clear separation of iron-sufficient (IS) and iron-deficient (ID) infants at 6 mo of age (A), which was not as pronounced by 12 mo, when the iron deficiency had resolved (B); n = 10.

In total, the serum concentrations of 413 known metabolites were measured (Supplemental Table 2). These identified metabolites (P < 0.05) were used in aggregate to inform affected pathways because groups of metabolites were more informative than any single metabolite after multiple hypothesis testing. In addition, a directed analysis of pathways markedly reduced the number of analyses and likelihood of inadvertent false discovery.

The ID monkeys had distinctly different concentrations of 40 metabolites (P 0.05) at 6 mo of age when compared with the age-matched IS infants (Table 1). By 12 mo of age, 48 metabolites were differentially expressed in the formerly ID monkeys relative to the 12-mo-old IS infants (Table 2). Several of these metabolites were of particular interest because they directly implicate involvement of the liver along with the many consequential events downstream of the hepatic response. At 6 mo of age, the ID monkeys had higher concentrations of 7α-hydroxy-3-oxo-4-cholestenoic acid (7-HOCA) and chenodeoxycholate (Figure 3A), but lower concentrations of 3 different bile acids, taurocholate, taurodeoxycholate, and tauroursodeoxycholate (Figure 3B, C, and D, respectively), relative to the age-matched IS monkeys. In addition, the ID monkeys had significantly higher concentrations of uridine and 5-methyluridine at this age, but other pyrimidine analogues and purine breakdown products were not significantly altered (Figure 4). Correlational analysis demonstrated a link between iron status and uridine, uracil, and 5-methyluracil concentrations on an individualized basis (Supplemental Figure 2). Iron deficiency was associated with changes in sn-glycero-3-phosphocholine, arachidonic acid, l-threonine, l-methionine, ribitol, chenodeoxycholate, betaine, glycine, decanoylcarnitine, and N-acetylneuraminic acid. IPA, a data analysis platform that evaluates fluctuations in individual molecules based on known metabolic pathways, was employed to interrogate these changes, and the result confirmed disturbed hepatic functioning, with a Fisher exact test P = 0.0000033 and an activation z-score of 2.077.

TABLE 1.

Serum metabolite ratios in ID relative to IS infant rhesus monkeys at 6 mo of age¹

Biochemical name	ID6mo/IS6mo	P value	q Value
1-Methylhistidine	1.800	0.004	0.125
10-Heptadecenoate (17:1n–7)	1.420	0.075	0.464
2-Hydroxydecanoic acid	1.810	0.014	0.238
3-Ethylphenylsulfate	0.690	0.026	0.301
3-Hydroxydecanoate	1.920	0.038	0.360
3-Hydroxyoctanoate	2.060	0.022	0.290
5-Hydroxymethyl-2-furoic acid	0.650	0.025	0.301
5-Methyluridine (ribothymidine)	1.440	0.010	0.205
6-Oxopiperidine-2-carboxylic acid	0.720	0.006	0.154
7-α-Hydroxy-3-oxo-4-cholestenoate (7-HOCA)	1.440	0.005	0.142
Acisoga	1.310	0.043	0.374
α-Hydroxyisocaproate	0.830	0.034	0.346
Arachidonate (20:4n–6)	1.300	0.046	0.382
Betaine	0.840	0.032	0.330
Butyryl carnitine	1.630	0.025	0.301
Caprate (10:0)	5.630	0.029	0.315
Carnitine	1.520	0.049	0.385
Chenodeoxycholate	3.150	0.019	0.265
Decanoylcarnitine	2.740	0.014	0.238
Dihomo-linolenate (20:3n–3 or n–6)	1.460	0.004	0.126
Docosadienoate (22:2n–6)	1.790	0.014	0.238
Glycerophosphorylcholine (GPC)	1.360	0.047	0.383
Glycine	0.790	0.018	0.265
Heme	0.390	0.051	0.393
Hypoxanthine	1.340	0.030	0.316
Isobutyrylglycine	0.640	0.003	0.105
Laurate (12:0)	1.930	0.022	0.290
Methionine	0.840	0.042	0.373
N-acetylisoleucine	0.740	0.026	0.301
N-acetylneuraminate	1.370	0.014	0.235
Palmitoylcarnitine	1.400	0.031	0.327
Phenol sulfate	0.500	0.009	0.191
Ribitol	1.220	0.042	0.373
Ribulose	1.770	0.007	0.171
Serotonin (5HT)	1.370	0.037	0.360
Taurocholate	0.300	0.011	0.206
Taurodeoxycholate	0.360	0.014	0.235
Tauroursodeoxycholate	0.470	0.037	0.360
Threonine	0.800	0.044	0.375
Uracil	1.480	0.046	0.382
Uridine	1.240	0.001	0.054
Xylitol	1.270	0.027	0.303

¹Significant metabolomic data displayed as normalized, imputed, ANOVA contrast ratios across experimental groups, with accompanying P value and q value for those ratios. Metabolites are shown with their common PubChem name. Values are fold change based on random forest classification and repeated-measures ANOVA after log transformation and imputation with minimum observed values for each compound; n  = 10. ID, iron-deficient; IS, iron-sufficient.

TABLE 2.

Serum metabolite ratios in ID relative to IS infant rhesus monkeys at 12 mo of age¹

Biochemical name	ID12mo/IS12mo	P value	q Value
1-Arachidonoylglycerophosphocholine	1.260	0.019	0.265
1-Docosahexaenoylglycerophosphocholine	1.290	0.029	0.315
1-Eicosatrienoylglycerophosphocholine	1.390	0.049	0.385
10-Heptadecenoate (17:1n–7)	1.870	0.008	0.178
10-Nonadecenoate (19:1n–9)	1.930	0.004	0.126
2-Aminoadipate	0.690	0.027	0.303
2-Hydroxyglutarate	0.680	0.029	0.315
2-Hydroxyhippurate (salicylurate)	0.700	0.018	0.261
2-Linoleoylglycerophosphocholine	1.340	0.049	0.385
2-Palmitoylglycerophosphocholine	1.360	0.037	0.360
3-Hydroxydecanoate	1.690	0.041	0.369
3-Ureidopropionate	0.730	0.007	0.167
Acetyl carnitine	0.660	0.009	0.200
Adipate	0.710	0.008	0.179
Betaine	0.830	0.025	0.301
Bilirubin (E,E)	1.390	0.024	0.301
Butyryl carnitine	0.620	0.024	0.301
Corticosterone	0.740	0.036	0.356
Cortisone	1.170	0.048	0.385
Dihomo-linoleate (20:2n–6)	1.790	0.003	0.108
Dihomo-linolenate (20:3n–3 or n–6)	1.410	0.003	0.108
Docosapentaenoate (n–3 DPA; 22:5n–3)	1.650	0.002	0.076
Docosapentaenoate (n–6 DPA; 22:5n–6)	1.560	0.002	0.091
Eicosapentaenoate (EPA; 20:5n–3)	1.420	0.004	0.126
Eicosenoate (20:1n–9 or n–11)	1.730	0.016	0.252
Erucate (22:1n–9)	1.480	0.016	0.253
γ-Glutamylalanine	0.740	0.010	0.205
Glutamine	0.880	0.044	0.378
Heme	0.200	0.028	0.307
Kynurenine	0.780	0.025	0.301
Laurate (12:0)	1.500	0.026	0.301
Linoleate (18:2n–6)	1.500	0.019	0.269
Linolenate [α or γ (18:3n–3 or n–6)]	1.640	0.030	0.315
Margarate (17:0)	1.600	0.004	0.117
Mead acid (20:3n–9)	1.310	0.045	0.381
Myristate (14:0)	1.520	0.020	0.278
Myristoleate (14:1n–5)	1.940	0.030	0.318
N-acetylthreonine	0.790	0.004	0.126
Nicotinamide	0.780	0.039	0.362
Nonadecanoate (19:0)	1.310	0.042	0.373
Palmitate (16:0)	1.490	0.017	0.260
Palmitoleate (16:1n–7)	1.980	0.012	0.216
Palmitoylcarnitine	1.310	0.041	0.369
Phenethylamine (isobar with 1-phenylethanamine)	0.800	0.036	0.352
Quinolinate	0.700	0.018	0.265
Stearate (18:0)	1.400	0.006	0.157
Stearidonate (18:4n–3)	1.420	0.049	0.385
Urobilinogen	0.780	0.027	0.303

¹Significant metabolomic data displayed as normalized, imputed, ANOVA contrast ratios across experimental groups, with accompanying P value and q value for those ratios. Metabolites are shown with their common PubChem name. Values are fold change based on random forest classification and repeated-measures ANOVA after log transformation and imputation with minimum observed values for each compound,  n  = 10. ID, iron-deficient; IS, iron-sufficient.

FIGURE 3.

Quantitative metabolomic measurements of chenodeoxycholate (A), taurocholate (B), taurodeoxycholate (C), and tauroursodeoxycholate (D) in rhesus monkeys that were iron-sufficient (IS) or iron-deficient (ID) at 6 mo of age prior to being fed solid food, and at 12 mo of age. Values are mean SEM, n = 10. Main effects of age, iron status, and age × iron status, P < 0.05, 2-factor ANOVA. Groups without a common letter differ, P < 0.05.

FIGURE 4.

Quantitative metabolomic measurements of uridine (A), 5-methyluridine (B), xanthine (C), and pseudouridine (D) in rhesus monkeys that were iron-sufficient (IS) or iron-deficient (ID) at 6 mo of age prior to being fed solid food, and at 12 mo of age. Values are mean SEM, n = 10. Main effects of age, iron status, and age × iron status, P < 0.05, 2-factor ANOVA. Groups without a common letter differ, P < 0.05.

Resolution of iron deficiency after weaning from the mother and the dietary transition to iron-containing solid foods resulted in a significant change for 20% of the measured metabolites between 6 and 12 mo in ID monkeys (P 0.05; Supplemental Table 2, columns 7, 13, and 14). The concentrations of several long-chain fatty acids were higher at 12 mo of age in the formerly ID monkeys relative to their concentrations at 6 mo (Supplemental Table 2, columns 7, 13, and 14). Conversely, the concentrations of several pregnenolone steroids and medium-chain fatty acids were lower at 12 mo relative to the concentrations at 6 mo (Supplemental Table 2, columns 7, 13, and 14). Developmental changes in metabolites were also observed in the IS monkeys, but the number of significantly altered metabolites was much smaller when compared with the ID infants over the same time frame. Only 8% of measured metabolites in the IS infants showed an age-related change between 6 and 12 mo (P 0.05) (Supplemental Table 2, columns 8, 15, and 16).

Discussion

Infantile iron deficiency in this nonhuman primate model indicates that there is a significant impact on the serum metabolomic profile and that many metabolite alterations persist even after the natural resolution of iron deficiency as measured by the traditional iron-sensitive hematological indices. Anemia, the current clinical standard for diagnosis and treatment of iron deficiency in infants (27), is an advanced state of iron deficiency evident in the blood, and tissue iron deficits are already present in the liver, kidneys, skeletal muscle, heart, and brain by the time a low Hgb concentration is detected (7, 8, 35). Therefore, our findings reaffirm the importance of early detection, and indicate that closer attention should be paid to the known prenatal and early postnatal risk factors, such as preterm birth and a more rapid postnatal growth rate, both of which predispose for iron deficiency. The difference in the extent of the maturational changes between 6 and 12 mo of age in the formerly ID monkeys suggests that although the bioavailability of iron in the solid food diet had enabled a correction of iron deficiency, the metabolic impact was still evident when compared with an infant who had remained IS throughout the first year of life. Several consensus reports have recommended that pediatricians consider earlier screening for iron deficiency; for example, at 6 mo rather than at 12 mo, in populations at higher risk for infantile iron deficiency (36, 37).

Many of the metabolomic changes found in the ID monkeys likely arose from altered function in iron-regulated physiological pathways, particularly in the liver, an iron-dependent organ affected before other tissues during iron deficiency (7). The liver serves many important body functions including production of lipoproteins, transferrin, complement, and glycoproteins, as well as having a role in lipogenesis, lipid metabolism, gluconeogenesis, bile acid synthesis, and oxidative degradation of toxins. Thus, it is not surprising that many of these processes were overtly perturbed in the ID state at 6 mo of age. However, the extent of the persistence of this metabolic perturbation that was still evident at 12 mo in the formerly ID monkeys is of greater concern. From a clinical perspective, it highlights the importance of early intervention in the form of iron supplementation. It is not known if these lingering effects partially reflect the adaptations needed to recover from iron deficiency while still sustaining rapid growth, or if there could be a more fundamental reprogramming of hepatocytes due to several months of low iron availability. Prior studies have demonstrated that there can be long-term injurious metabolic consequences of severe malnutrition during infancy (38, 39), but the literature is less consistent with regard to concerns about undernutrition when it involves a single nutrient. In carefully controlled experiments with animal models of iron deficiency it is possible to demonstrate a sustained impact on the brain and other organs that persists into adulthood (13, 40, 41). Numerous reports indicate that early-life iron deficiency causes long-lasting neurobehavioral and cognitive deficits that persist into early adulthood, indicating some level of permanent alterations in brain function (42–45).

Bile acids are produced by the liver, and it has been known for a long time that they are important for iron solubility and absorption (46). Evidence for higher concentrations of bile acids in the ID monkeys could indicate a compensatory response to maximize intestinal absorption of iron in a more acidic milieu (47). In addition, whereas iron absorption normally occurs primarily in the duodenum and upper jejunum, when conjugated with bile salts, iron absorption can be extended into the lower small intestine. This relation might also explain why 7-HOCA returned to the normal range after restoration of iron concentrations in the older infants. Plasma concentrations of 7-HOCA reflect cytochrome p450 7A1 (CYP7A1) activity in the liver, which synthesizes the precursor of 7-HOCA, 7-α-hydroxycholesterol (48). Bile acids are synthesized in the liver from cholic and chenodeoxycholic acids via cytochrome P450–mediated oxidation of cholesterol. Hepatocytes must conjugate either glycine or taurine to the precursor cholic or chenodeoxycholic acid to form a total of 8 primary bile salts. Disrupted bile acid production has also been described in iron-deficient anemic rats (49), and the current study has now replicated these conclusions in a developmental model of infantile iron deficiency in monkeys, which have a developmental time course more similar to human infants.

Uridine can be synthesized from uracil, and the concentrations of both compounds, along with that of 5-methyluridine, were elevated in ID monkeys, yet other pyrimidine analogues, such as pseudouridine, were not altered. It does not appear that this increase in uridine, uracil, and 5-methyluridine is attributable to excessive degradation of purines and pyrimidines in general because concentrations of xanthine, a purine breakdown product, were unchanged with iron deficiency. The liver is the primary source of uridine, although fat cells can take over this function during periods of fasting (50). Uridine is required not only for RNA synthesis, but also for the proper processing of galactose (51). In the fed state, the liver secretes uridine into bile, where it is ultimately transferred into the small intestine to improve absorption of nutrients. Injection of uridine causes a reduction of body temperature in rats, indicating that elevated uridine concentrations provide a mechanism for conserving energy through lowered body temperature and metabolic rate (50). This finding is in keeping with the fact that abnormal thermoregulation and lower body temperature are typically observed in anemic humans and rats (52). The effects of iron deficiency on thermoregulation can also be through the hypometabolic state caused by reduced thyroid hormone synthesis due to lower concentrations of the iron-dependent deiodinase 2 enzyme (53–55). The developing infant has significant energy requirements, notably to support the maturation of the brain and liver, and reduced metabolism and altered β-oxidation would likely impact many substrates required for proper development. Because this effect was not observed across all nucleosides, pyrimidines, purines, or their breakdown products, it appears to be specifically regulated (56, 57). Thus, it is possible that the metabolic downregulation should be viewed as a physiologically adaptive response, in order to better match metabolic rate with substrate availability.

Iron deficiency in rats also increases hepatic lipogenesis, which leads to steatosis and triacylglycerol accumulation (58). The elevations we observed in serum fatty acids in 12-mo-old infants after recovery from iron deficiency could be due to an accumulation of fatty acids in response to the restoration of enzyme function and a decrease in lipogenesis. 3-Hydroxybutyrate (3-HBA) would be indicative of excess acetyl CoA production, which usually is a result of increases in fatty acid oxidation (59). An increase in serum fatty acids, coupled with a decrease in 3-HBA, could indicate changes in energy sourcing and decreased fatty acid β-oxidation during the period of recovery from iron deficiency. In the developing brains of rats, mice, and rabbits, iron deficiency results in reduced myelination due to a shift away from phospholipid production in favor of cholesterol, and reduced functionality of oligodendrocytes (60, 61). In anemic rats, there is also a shift toward increased concentrations of cholesterol esters and free cholesterol (62), but reduced phospholipid concentrations (63) in serum, consistent with the results from our monkey model of iron deficiency.

It should be acknowledged that a limitation of this type of metabolomic profiling is that it generates a very large number of end points, which can potentially lead to the false discovery of differences between groups. The primary focus was between just 2 conditions: the metabolite differences between ID and IS infants during peak iron deficit, which occurs at 6 mo of age in this animal model (9, 12, 28). To further address the issue of multiple comparison, the meaningfulness of differences in individual metabolites was considered in a modeling of pathways, which markedly reduced the number of potential outcomes. Finally, the biological significance of the initial differences observed at 6 mo of age was validated and considered in the context of whether metabolic pathways in 2 groups of infants were found to still be different when re-evaluated at 1 y of age.

In summary, MS analysis of serum provided a sensitive platform to interrogate the metabolomic profile associated with iron deficiency in a manner that can be directly applied to human infants in the clinical setting. A number of key metabolic pathways were perturbed during the period of iron deficiency and remained altered for months following its natural resolution. These findings highlight the critical importance of early detection and the need for refining more tests that can screen for collateral effects of iron deficiency beyond the traditional hematological measures and reliance on Hgb as the primary indicator of iron deficiency. Our previous research using this nonhuman primate model documented parallel effects on brain energetics and metabolism (12, 13). Our current efforts are devoted to validating sensitive bioindicators of brain health that can be acquired from blood, to identify at-risk infants at very early stages of iron deficiency. It is also important to identify the nature of the cellular reprogramming that can occur in iron-dependent tissue when impacted by iron deficiency for many months, especially in the liver and brain. Metabolomic profiling can also help guide the optimal timing for targeted therapeutic modalities to address the downstream effects of iron deficiency and to track any residual metabolic effects that may linger after iron treatment.

Notes

This work was supported by grants NIH R01HD089989, NIH R01HD080201, NIH R01HD057064, and NIH R01HD39386. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Author disclosures: The authors report no conflicts of interest.

Supplemental Tables 1 and 2 and Supplemental Figures 1 and 2 are available from the “Supplementary data” link in the online posting of the article and from the same link in the online table of contents at https://academic.oup.com/jn/.

Abbreviations used: CCA, canonical correlation analysis; CSF, cerebrospinal fluid; HCT, hematocrit; Hgb, hemoglobin; ID, iron-deficient; IPA, Ingenuity Pathway Analysis; IS, iron-sufficient; MCH, mean corpuscular hemoglobin; MCHC, mean corpuscular hemoglobin concentration; MCV, mean corpuscular volume; RDW, red blood cell distribution width; TSAT, transferrin saturation; ZnPP/H, zinc protoporphyrin/heme; 3-HBA, 3-hydroxybutyrate; 7-HOCA, 7α-hydroxy-3-oxo-4-cholestenoic acid.