The purpose of this study was to determine the impact of dietary maternal iron deficiency (ID) on fetal development, with the main focus on the identification of critical periods of gestation when the developing CNS is most vulnerable to maternal ID. The animal model we established was intended to mimic the extremely prevalent human condition of marginal maternal ID that occurs in an estimated 30 to 50 percent of pregnancies worldwide. After careful titration of the iron concentration in the diet, we were able to establish a model where the pregnant dam was rendered iron deficient but did not develop severe anemia during pregnancy.
The analysis of the most prolonged form of diet restriction, in which iron levels were already reduced prior to the onset of pregnancy led to some unexpected findings. For example, we found that despite the fact that the dams received the iron-deficient diet not only prior to conception but during the entire pregnancy, the dams did not develop severe anemia that had an impact on hematocrit levels, which remained in the normal range. However, maternal serum iron concentrations decreased continuously throughout pregnancy, confirming iron depletion. Upon more detailed examination of the maternal blood via HESKA Hema True Veterinary Hematology System, we observed minimal reductions in the measurement of mean corpuscular volume and hemoglobin compared to control values. Taken together, these results are consistent with ID and the absence of severe anemia.
This dietary model thus mimics a situation that would not cause any level of clinical concern in the human population, as the absence of severe anemia would not prompt any immediate intervention and is likely to remain unnoticed. We believe that this is an important aspect of our study as it underscores the need for monitoring the status of iron beyond the level of anemia.
Iron restriction to the dam two weeks prior to conception was associated with lower iron concentrations in embryonic tissues as early as embryonic day 15 (E15). The embryos maintained a very similar low iron level throughout gestation without an appreciable increase in serum iron concentrations suggesting that maternal iron stores, that apparently were not severely depleted, did not compensate for the continuously increased demand for iron by the embryo. Our observation seems to argue against the often proposed notion stating that the developing embryo is partly protected from the maternal ID through compensatory changes in the iron transport mechanisms of the placenta, which minimize the impact of ID in the fetus 
. However, our data is consistent with (i) recently published data, suggesting that gestational ID might have a restrictive effect on the constitution of adequate fetal iron stores, 
and with (ii) the observation by Naghii et al., indicating that fetal brain tissue can be iron depleted even in the absence of anemia in the pregnant mother 
. Furthermore, studies from the Georgieff laboratory have shown that a transgenic model of hippocampus-specific iron deficiency without any signs of anemia also resulted in altered neuronal development as well as significant cognitive dysfunctions (reduced spatial recognition memory performance and procedural memory) 
. Wu et al. have also shown that a short period of perinatal iron deficiency results in altered associative behavior in adult rats without anemia 
. These studies support our notion that anemia is not necessarily a defining predictor of neural impairments that are associated with iron deficiency.
Our model of gestational ID was not only associated with persistently lower body weights as early as E15 but was also associated with functional impairments as determined by the Auditory Brainstem Responses (ABR) analysis of neural conduction velocity in the offspring. The increase in P2-P1 interpeak latencies without hearing threshold changes suggested that, at least in part, impaired myelination could be one explanation for our findings. While disruptions in myelination have been demonstrated in rodents with prenatal and lactational ID 
and hypomyelination is also recognized as a defining feature associated to perinatal ID in prematurely born infants 
, the greater increase in latency changes seen also by analyzing P4-P1 interpeak latencies (which reflects the entire brainstem) is consistent with findings in the literature suggesting that neuronal components could also be affected by iron deficiency.
As we established a functional criterion for the “diagnosis” of neural impairment that could clearly be associated with the maternal dietary intake, it was possible to design a timed dietary restriction study that cannot be performed in humans but is critical for understanding the window of vulnerability. We generated maternal dietary groups that were defined by the onset of iron restriction solely based on food source. Using the ABR test as a functional readout of the neural CNS integrity of the offspring, we determined that the maternal exposure to an iron-deficient diet either prior to conception, at the start of first trimester, or at the onset of second trimester had a significant negative impact on the ABR latency response in the offspring, placing the window of vulnerability for the fetus in the first two trimesters of gestation. Interestingly, the negative CNS effects on the fetus were more severe when the ID was initiated at the beginning of pregnancy compared to ID initiated prior to the onset of pregnancy. An important result of our research was that the offspring from pregnant rats receiving the ID diet at the beginning of the third trimester exhibited normal auditory nerve conduction velocities as reflected by ABR recordings despite IDA being present in all experimental rats at the time of testing. This result might shed some light on seemingly contradictory results found in humans; some investigators reported no alterations of ABR latencies in ID children with or without anemia 
, while others found that IDA in 6 month old babies produced abnormally prolonged ABR latencies that persisted years after the children's iron status was corrected with iron therapy 
. Our study indicates that there is a precise window of vulnerability during which ID can affect the fetal CNS leading to postnatal neural impairments. The lack of defect seen in some studies could be a consequence of the insult occurring outside the window of vulnerability, rather than a perceived general insensitivity of newborns to iron restriction. Without the establishment of the maternal iron status during pregnancy or at least cord blood iron levels, it cannot be resolved whether the initiation of ID and the possible onset of a later apparent pathology happened in utero
. Therefore, any conclusions about the importance and/or need of pre- and postnatal iron supplementation have to be made with caution.
The window of vulnerability, which we have defined in our studies, also offers an important clue for the underlying cellular impairment and seems to be consistent with our previous work on the study of cellular targets of ID 
. As illustrated in , ID during embryonic CNS development seems to affect neural precursor cell populations leading to an imbalance of specific precursor populations, which might be ultimately responsible for reduced oligodendrocyte numbers, abnormal myelination, and neural pathologies. We have shown previously that the differentiation and proliferation of embryonic glial precursor cells can be modulated by exogenous iron concentrations 
. Interestingly, the timing of the generation, expansion and differentiation of these precursor cells coincides with the window of vulnerability we have defined in our diet regimens. In the rat embryo, gliogenesis starts around day E13.5 with the generation of embryonic glial restricted precursor cells (GRPs), which shortly before birth give rise to oligodendrocyte precursors (OPCs) that persist postnatally 
. Even slight alterations of proliferation and/or of differentiation abilities of precursor cells (GRPs and OPCs) occurring before the time of oligodendrocyte generation could translate into a myelination defect. The notion of early embryonic cells being especially responsive to suboptimal iron levels is also supported by studies from Badarocco et al., whose interesting experiments using intracranial apotransferrin injections suggest that iron may affect oligodendrocyte development at early stages of embryogenesis rather than during late development 
. The defects that are associated with iron deficiency are, however, unlikely to be restricted to the glial population. As shown in , neuronal progenitor cells develop in a similar time window and could equally well be targets of iron deficiency. The cognitive impairment seen in the offspring is most likely a combination of an impact on both neuronal and glial embryonic progenitor cell pools.
Proposed model displaying the impact of gestational iron restriction on characterized neural cell populations along a developmental time line.
The absence of a functional defect in late onset ID suggests that, if the diet is introduced at embryonic day 14 (red arrow) and maintained, the time point of iron depletion is past the period when the most vulnerable embryonic populations expand and generate their progeny. Under these conditions, the impact of ID on glial cell generation and/or neuronal maturation does not seem to be severe enough to result in detectable auditory dysfunction.
It should be pointed out that despite the lack of an ABR latency defect in our study at the late gestational time point (ID E14
), we cannot exclude the presence of other abnormalities. For instance, a number of studies in rodent and primate models of late gestational and postnatal ID reported significant and persistent behavioral abnormalities and neurochemical disturbances 
. As development occurs along a temporal continuum, it would be expected that deficiencies occurring at different times cause different types of outcomes.
The timed maternal iron restriction coupled with a persistently deficient diet used in our model, led to another surprising finding that sheds some new light on the role of anemia in iron deficiency. Due to the associated fatigue, muscle weakness, and dyspnea during exertion, anemia is considered a confounding factor in various behavioral tests and raises criticism on the accuracy of the results and their interpretation. The ABR electrophysiological test we employed in the evaluation of neurodevelopmental impairments induced by ID seems not to be influenced by anemia. As shown in , all offspring of our different feeding groups developed clinically relevant anemia, yet only three out of four groups showed functional impairment in ABR patterns. This observation strongly suggests that anemia is not a determinant factor of pathological ABR patterns recorded in iron deficient animals and allows dissociation of the impairment caused by ID from effects that might be primarily due to anemia. As ABR analysis can be conducted in multiple species, including humans, this type of analysis may provide a particularly useful noninvasive tool for examining the effects of micronutrient deficiencies on CNS development.
While IDA was not required to cause ABR defects, the occurrence of IDA did, however, severely impact the survival rate of the offspring. The litter survival rate was approximately 75% in the four experimental groups compared to the control groups. This decrease in survival rate may lead to an unintentional sample population bias of “survivors” at P45 when the functional measurements are conducted. While we cannot completely rule out that the surviving animals are on some level more robust that their littermates, it is important to note that we cannot correlate survival with absence or presence of ABR defects. The ID -2wk, ID E0 and ID E7 groups had as many animals that died during the experiments as the ID E14 group yet none of the surviving animals in the last group showed an ABR defect, while all the animals in the other three groups showed an ABR defect of various severities.
Another novel observation in our study is the time-specific presence of thermoregulatory deficits in the offspring from mothers who were iron restricted two weeks prior to conception. The three other feeding groups did not develop such defects. As ID has sometimes been linked to poor thyroid function 
, which in turn can result in a thermoregulatory dysfunction 
, the decreased core body temperature in the most severe feeding regimen (ID -2wks)
might involve a thyroid component. However, based on our data, we do not think that thyroid involvement is critical for the development of the ABR dysfunction, as the ID E0
and the ID E7
groups have significant ABR dysfunctions but normal core body temperatures.
In summary, the analysis of developmental defects that are associated with exposure to iron restriction during different phases of fetal development emphasizes the influence of the precise window of vulnerability and the associated duration and magnitude of ID that is exerted on the developing fetal CNS. The prolonged absolute and interpeak latencies revealed by the comprehensive ABR analysis in ID animals could have important clinical implications. If ID at an early age causes hypomyelination or delayed neural maturation along the auditory pathway altering the properties of the auditory message and its transmission through the brainstem structures, it is possible that such a dysfunction could have functional significance for language acquisition and the development of other higher cognitive and emotional functions.
In lieu of this information, our data not only reiterates the importance of iron supplementation as part of prenatal and gestational IDA therapeutic regimen but also as a preventive strategy. Identifying the windows of vulnerability when specifically vulnerable cell populations are generated during embryonic development will be critical in delineating the restricted window of therapeutic opportunity when iron supplementation could correct or prevent the functional deficit.