Sex differences in adult social, cognitive, and affective behavioral deficits following neonatal phlebotomy‐induced anemia in mice

Abstract Introduction Anemia is common in prematurely born infants due to blood loss resulting from frequent phlebotomies and may contribute to their neurobehavioral deficits. Preclinical models of phlebotomy‐induced anemia (PIA) have revealed metabolic and genomic changes in multiple brain structures of young mice, yet the impact of neonatal PIA on early‐life and adult behavior has not been assessed. Methods The present study employed a range of behavioral measures in phlebotomized anemic neonatal mice to investigate short‐ and long‐term neurodevelopmental effects. PIA from postnatal (P) days 3 to 14 caused sex‐specific changes in social behavior, novelty preference, and anxiety at P17 that persisted into adulthood. Results Our preclinical model suggests that PIA may contribute to acute and long‐term behavioral and affective deficits and warrants further substantiation of the observed behavioral phenomena in larger samples. Conclusions We conclude that this model is a useful tool for beginning to better understand the lasting effect that early‐life PIA might have on the developing brain. The differential impact of PIA on male and female subjects warrants further exploration for the development of appropriately targeted interventions.

assessed neurodevelopmental outcomes as a function of the degree of neonatal anemia were unable to determine the degree of anemia that is safe for the developing brain. The Iowa trial (Bell, 2005) showed increased risk of severe hemorrhage and the PINT trial (Kirpalani et al., 2006) showed poorer overall performance on standardized developmental testing in more anemic infants. Long-term follow-up of infants from the Iowa trial demonstrated poorer longterm outcome only in males, whereas females with lower hematocrits had better outcomes than those randomized to higher hematocrits (McCoy et al., 2011;Nopoulos et al., 2011). Transfusion is a potential risk factor to the developing organs, including the brain (Benavides et al., 2019). Because of the apparent risks of transfusions, fewer infants are being transfused and more infants are allowed to maintain lower hematocrits without consideration of potential neurometabolic (Wallin et al., 2015(Wallin et al., , 2017 and behavioral consequences. Anemia is a risk to the developing brain because it compromises the delivery of key metabolic and energetic substrates, such as iron and oxygen. In preterm infants, much of anemia is caused by significant phlebotomy-induced blood loss (Carroll & Widness, 2012;Widness, 2008). Studies indicate that inadequate iron and oxygen status early in life has a substantial negative effect on the development of the brain as well as other organ systems (Lozoff & Georgieff, 2006). In particular, early-life iron deficiency is predictive of subsequent mental health complications, including severe psychopathology (Barks, Fretham, Georgieff, & Tran, 2018;Insel, Schaefer, McKeague, Susser, & Brown, 2008;Lozoff & Georgieff, 2006;Schmidt, Tancredi, Krakowiak, Hansen, & Ozonoff, 2014).
The objective of the present study was to examine the acute and long-term behavioral effects of early-life phlebotomy-induced anemia (PIA) in an established mouse model (Wallin et al., 2015(Wallin et al., , 2017. In that model, PIA with a hematocrit of 25% increased hippocampal lactate concentrations by 60% and reduced expression of critical synaptic plasticity genes, for example, BDNF, in the neonatal period (Wallin et al., 2015(Wallin et al., , 2017. In this study, we employed a set of behavioral paradigms with the same model to assess a broad range of functional domains (social behavior, anxiety, cognitive performance) in phlebotomized male and female mice at postnatal (P) day 17 and again in adulthood starting at P65. Further, we tested the effects of two postphlebotomy hematocrit (Hct) target levels (25% Hct and 18% Hct) on these measures to establish whether a dose-dependent effect was present. The chosen Hct values were selected based on human data (Bell, 2005;Kirpalani et al., 2006), using the same percent reduction relative to control (nonphlebotomized) hematocrits utilized in the human studies. Specifically, while physiologic anemia and neonatal Hct in mice are similar to those observed in humans, the mice used in the present model had a lower starting Hct (38%) than values recorded in the human (50%). Thus, in addition to achieving 25% Hct following phlebotomy, a second target Hct level of 18% was included to reflect the same percent separation from nonphlebotomized controls as is present in humans. The study design also allowed for dose-response analysis as a function of degree of anemia.

| Animal preparation
The study was conducted with the approval of the Institutional Animal Care and Use Committee at the University of Minnesota.
Wild-type C57/BL6 animals were used for all the experiments in the study. Pregnant and lactating dams were fed standard chow containing ~200 ppm of iron (Envigo; Indianapolis, IN) and were given access to food and water ad libitum. The animals were maintained on a 14-hr light/10-hr dark cycle with humidity and environmental temperature controlled as per the institution's Research Animal Resources guidelines. Litters were culled to 8-9 pups at P2 or, in the case of smaller litters, foster pups were added to maintain similar litter size. This was done to reduce variability of food access and growth rates resulting in disproportionate weight gain. The pups along with the lactating dams were transferred to the behavioral testing facility at P14 and were allowed to acclimatize to their new environment for 2-3 days before the beginning of testing. Mice tested at postnatal day 17 will be referred to as "P17" mice, followed by their Hct status in the remainder of the study, while mice tested after postnatal day 65 will be referred to as "P65" mice, followed by prior Hct status. A breakdown of the number of animals of each sex included in each of the treatment groups is provided separately for every task below.

| Phlebotomy-induced anemia
Neonatal mice were phlebotomized from postnatal (P) day 3 to P13 via facial vein venipuncture using a micropipette as described previously (Wallin et al., 2015(Wallin et al., , 2017. Pups were weighed daily to determine the quantity of blood to be drawn. For both 25% hematocrit and 18% hematocrit groups, blood was drawn twice daily at 5.25 µL/g until goal hematocrit levels were reached (P7-P8 for the 25% group and P10-P11 for the 18% group). The animals were bled daily at 3.5 µl/g thereafter to maintain the target hematocrit levels. The hematocrit was measured daily by centrifugation of microhematocrit collection tubes at 10,000 rpm for 5 min and quantified using a hematocrit card reader. Control pups received a nonphlebotomizing needle prick to the nape of the neck on the same schedule as phlebotomized pups.
They were handled similarly, including time spent away from the dams, to the phlebotomized pups in order to minimize variability in stress responses. Each litter comprised pups from both control and phlebotomized groups.

| Three-chambered social approach task
We tested 19 nonbled controls (males = 8), 21 25% Hct mice (males = 7), and 19 18% Hct mice (males = 9) at P17, and 13 controls (males = 5), 22 25% Hct mice (males = 8), and 21 18% Hct mice (males = 9) at P65. The three-chambered social approach task was used to assess sociability and social approach behavior 25 . In this experiment, animals underwent three 10-min phases within an arena (20 × 25 × 45 cm) partitioned into three chambers. In the first (habituation) phase, individual mice were allowed to explore the arena freely. In the second (sociability) phase, an age-and sexmatched conspecific was enclosed under a small wire mesh cage (cup) within one chamber (hereafter referred to as the "mouse" chamber), while the other chamber contained an empty cup. The side of the apparatus containing the conspecific was randomized. In the third phase of the task, a novel conspecific was enclosed under a wire mesh cup on the opposite side of the chamber. For each phase, the time spent within each chamber and investigating each wire cage were scored by a genotype-naïve (blind) experimenter. The ratio of time spent in the novel versus familiar chamber was calculated during the novelty phase.

| Novel object recognition (NOR)
NOR test procedures were similar to those described previously by our laboratory 26 and others 27 . A total of 19 nonbled controls (males = 12), 21 25% Hct mice (males = 13), and 20 18% Hct mice (males = 9) were tested in this task. Both acquisition and test phases were videorecorded. Prior to each phase, animals were habituated to the test chamber for 1 min. Mice were then removed, objects were placed in the chamber, and the animals were returned to the chamber. During the acquisition phase, two identical objects were placed approximately 5 cm away from the opposite walls of the chamber and mice were allowed to explore freely for 5 min. At the end of the trial, mice were removed and then returned to the chamber for the test phase 15 min later. During the test phase, the familiar object from the acquisition phase was placed in one location within the chamber and a novel object was placed in the opposite location.
Animals were placed in the test chamber containing both objects for five minutes. Upon completion of the test phase, mice were returned to their home cages. The test chamber was thoroughly wiped down with 70% alcohol, and objects were cleaned with diluted bleach between sessions.
The test phase made use of three different pairs of novel and familiar objects (for a total of 6 objects), which varied in color, material, texture, and shape. The location of the novel object in the test chamber, as well as test order and object pair, was counterbalanced across conditions. Object investigation during testing was scored by trained group-naive raters. During the acquisition phase of the task, time spent exploring either of the two identical objects placed in the testing apparatus was recorded. The total duration of investigation was calculated for each mouse (exploration score). During the test phase, a normalized novelty score was calculated for each animal by dividing the difference in investigation times for the novel and familiar objects by the time spent investigating the familiar object. Preference for the novel object (and memory for the familiar object) was defined as time spent investigating the unfamiliar object compared to chance during the test phase.

| Statistical analysis
Data were analyzed using GraphPad Prism. To assess the effect of sex and treatment on behavior, we used 2-way ANOVA and performed Tukey's and Sidak's multiple comparisons tests.

| Three-chambered social approach task
To test whether sex and treatment (hematocrit status) had an impact on behavior in this task and to examine the possible interaction between the two, we performed a 2-way ANOVA, with sex and treatment as factors (Figure 1). Among P17 animals, both treatment

| Novel object recognition
Preference for the novel or familiar object was measured as time spent exploring (sniffing, touching, climbing on) and facing the object. A novelty score was calculated for each animal in each group by dividing the difference between time spent investigating novel

| D ISCUSS I ON
Whether anemia without the neurodevelopmental risk of red cell transfusion (Benavides et al., 2019) is sufficient to compromise neurodevelopment, and whether deficits associated with anemia are short-lived or lasting, is poorly understood. Here, we addressed these questions in a validated preclinical model of phlebotomyinduced neonatal anemia (Wallin et al., 2015(Wallin et al., , 2017. Two Hct target concentrations (25% and 18%) allowed us to test the effect of PIA on cognitive, social, and affective functioning in a dose-dependent manner. These concentrations were chosen to be commensurate with those seen in human preterm infants after adjusting for lower starting Hcts in the mouse. We examined whether PIA produced acute effects on behavior and whether observed differences persisted into adulthood. We also assessed whether PIA produced different acute and long-term effects on males and females.
Previous studies using this neonatal PIA model have found evidence for brain hypoxia, as indicated by elevated VEGF expression in the hippocampus. Additionally, these studies have revealed a 40% reduction in total brain iron and an increase in hippocampal transferrin receptor-1 expression, confirming the presence of brain iron deficiency (Wallin et al., 2015(Wallin et al., , 2017. This model is also characterized by brain acidosis, evidenced by a 60% increase in hippocampal lactate concentration, and altered expression of hippocampal synaptic plasticity genes, including BDNF. Consistent with these findings of physiological dysregulation, in this study we found evidence consistent with impaired hippocampal function in the performance of PIA animals on the NOR task. The F I G U R E 2 Confidence intervals based on post hoc tests of P17 and P65 three-chambered sociability test data. Panels a,b: sociability (a) and social novelty (b) in P17 mice. Panels c-d: sociability (c) and social novelty (d) in P65 mice performance of PIA mice on this task was related to early Hct status, in that more severely anemic 18% Hct mice spent less time investigating the novel object, while 25% Hct mice showed a more moderate impairment. The observed deficits in novelty preference persisted into adulthood despite resolution of PIA, suggesting that anemia-induced changes to hippocampal integrity in early development have a lifelong effect on behavioral outcomes and that the severity of these outcomes is related to severity of anemia. Given the aforementioned tendency to decrease the number of transfusions in cases of neonatal anemia in NICUs, our findings highlight the possible long-term risks of prolonged exposure to increasingly severe anemia in preterm infants.
Beyond neurocognitive deficits, growing evidence demonstrates that preterm infants are at a markedly elevated risk for developing sociocognitive abnormalities including autism spectrum disorder (ASD) (Agarwal et al., 2018;Johnson et al., 2010;Limperopoulos et al., 2008;Mahoney, Minter, Burch, & Stapel-Wax, 2013). Although the neural mechanisms implicated in the etiology of sociocommunicative deficits and ASD remain incompletely understood, known risk factors include male gender, lower birthweight, gestational age, and infection (Limperopoulos et al., 2008;Ng, de Montigny, Ofner, & Do, 2017). Our study newly implicates neonatal anemia as another risk factor in the development of aberrant social behavior. In the PIA model, this included aversion to novel conspecifics and preference for nonsocial environments. An impairment in sociability and social novelty preference in the three-chambered social approach task was shown suggested a relationship between neonatal anemia and neurobehavioral deficits. Our results reveal that early (P17-P21) changes in social behavior impact males and females in both phlebotomized groups, though they were most pronounced in 25% Hct males, and that abnormalities persisted into adulthood. Both sociability and social novelty were deficient in 25% and 18% Hct animals at P17-P21 and did not reach control levels at P65. Notably, at P65, 25% Hct males performed better than 18% Hct males in the sociability portion of the task, and the latter spent the least amount of time in the compartment containing a conspecific versus the empty compartment. Similarly, 18% Hct males (P65) exhibited strong social novelty aversion; indeed, we observed a graded effect of treatment on performance in the social novelty phase of the experiment among male mice. Our results suggest a stronger impact of treatment in males on performance in the three-chambered sociability test, a result consistent with findings in the clinical literature.
Iron deficiency causes substantial changes in monoamine metabolism throughout the brain, with particularly large effects on the dopamine system (Beard, 2003;Chen, Beard, & Jones, 1995;Lozoff, 2011;Lozoff & Georgieff, 2006;Unger et al., 2012;Yehuda & Youdim, 1989;Youdim & Green, 1978). Such changes may explain iron-deficient infants' heightened risk for anxiety, as indicated by increased fear, wariness, and hesitancy (for a detailed review, see Lozoff, 2011). Previous preclinical studies utilizing the open field test to measure anxiety-associated behaviors in iron-deficient rats have reported subtle changes in indexes of arousal, such as defecation, as well as diminished rearing behaviors Felt & Lozoff, 1996). Others have reported a relationship between iron deficiency and elevated anxiety-like behaviors in the EPM (Li et al., 2011), although the literature linking iron deficiency and anxiety offers mixed findings (Beard, Erikson, & Jones, 2002;Eseh & Zimmerberg, 2005;Gewirtz, Hamilton, Babu, Wobken, & Georgieff, 2008) Asterisks denote a main effect of sex (F(1, 55) = 11.28, p = .0014), treatment (F(2, 55) = 6.366, p = .033), and a significant sex × treatment interaction (F(2, 55) = 6.549, p = .0028). Tukey HSD analyses showed that 25% Hct female mice spent significantly less time in the periphery of the open field (M = 318.807, SD = 215.361, p < 000.1). Panels c-d: confidence intervals based on post hoc tests of P17 and P65 data, respectively. "Mean differences" (x-axis) reflects the differences between means. *p < .05; **p < .01; ***p < .001; ****p < .0001. Means ± SEM Preclinical models of neonatal anemia due to dietary iron deficiency result in acute neurodevelopmental and behavioral morbidity Pinero, Li, Connor, & Beard, 2000;Ward et al., 2007). In light of this, the evidence presented here for compromised cognitive, social, and affective functioning early during development was expected. More surprising, however, was the enduring nature of these effects. The persistence of deficits into adulthood across several domains suggests that the functional integrity of multiple brain structures is not fully restored despite iron repletion. Such effects are concerning from a clinical standpoint in view of our limited understanding of the sustained effects of neonatal anemia on the developing brain and the potential societal costs of lifelong morbidity.
Two compatible theoretical explanations can account for the persistence of these behavioral changes. According to the critical period theory, regional brain structures undergo development during a critical period, disruption of which results in lasting structural and associated functional changes. Such effects have been observed as a result of fetal/neonatal iron deficiency and iron deficiency anemia (Brunette, Tran, Wobken, Carlson, & Georgieff, 2010;Fretham et al., 2012;Jorgenson, Wobken, & Georgieff, 2003;Schmidt, Waldow, Grove, Salinas, & Georgieff, 2007). The epigenetic theory posits that environmental factors, such as nutritional deficits, can alter the regulation of synaptic plasticity genes, both during the period of deficiency and into adulthood, thereby altering the function of the adult brain. Neonatal iron deficiency anemia has been shown to induce such effects on targeted genes (e.g., BDNF) and also on networks of genes that are associated with psychopathologies such as schizophrenia and autism (Tran, Kennedy, Lien, Simmons, & Georgieff, 2014;Tran et al., 2016). The effects of such changes in the adult hippocampus have been linked to abnormalities in NOR performance following early-life ID anemia .
Our data reveal that PIA differentially affects neurobehavioral outcomes in males and females. Although a similar conclusion has been suggested in clinical reports, it has been difficult to disentangle the effects of anemia from other confounds, including its treatment with red blood cell transfusions, which are thought to be F I G U R E 4 P65 elevated plus maze behavioral data from 19 nonphlebotomized controls (males = 8), 22 25% Hct mice (males = 8), and 22 18% Hct mice (males = 10). Panel a: A 2-way ANOVA revealed a main effect of sex (F(1, 57) = 43.18, p < .0001) and treatment (F(2, 57) = 7.239, p = .0016), and a significant sex × treatment interaction (F(2, 57) = 11.50, p < .0001), denoted by ****. Panel b: confidence intervals based on post hoc tests of P65 data. "Mean differences" (x-axis) reflects the differences between means. *p < .05; **p < .01; ***p < .001; ****p < .0001. Means ± SEM pro-inflammatory. While the mechanism driving these differences has yet to be uncovered, it has been proposed that female preterm infants show greater sensitivity to pro-inflammatory events (Nopoulos et al., 2011 (though we note that the results of this study should be considered with caution due to poor follow-up rates); Benavides et al., 2019). Indeed, the finding that females with higher Hct levels due to more numerous red cell transfusions harbored the most significant structural brain abnormalities relative to all other groups in this study seems to support this interpretation (but see a note of caution above). However, anemia itself also has a pro-inflammatory component, as demonstrated in preclinical models (Arthur et al., 2019). It is thus possible that anemia alone is sufficient to produce a strong pro-inflammatory response in female pups and that such a response would be commensurate with anemia severity. Since tissue cytokines were not measured in the current set of experiments, the interrogation of the putative relationship between anemia status and inflammatory response remains an objective of future work.

| CON CLUS ION
Preterm infants are at increased risk for cognitive, affective, and social deficits. A common practice in NICUs is to order multiple phlebotomies for laboratory testing, resulting in neonatal anemia.

CO N FLI C T O F I NTE R E S T S TATE M E NT
The authors declare no conflicts of interest.

ACK N OWLED G M ENTS
This work was funded by NIH grants P01-HL046925 and R01-HL138543.

AUTH O R CO NTR I B UTI O N S
TMM conducted behavioral experiments and data analyses and wrote the manuscript. GS performed phlebotomies, maintained the breeding colonies, and edited the manuscript. TAG, JCG, and MKG provided revisions to the manuscript, guided experimental design, and assisted with edits of the text and graphs.

PE E R R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1002/brb3.1780.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data are available upon reasonable request.