O-GlcNAc transferase (OGT) as a placental biomarkerof maternal stress and reprogramming of CNS genetranscription in developmentChristopher L. Howerton, Christopher P. Morgan, David B. Fischer, and Tracy L. Bale1

Department of Animal Biology, School of Veterinary Medicine, University of Pennsylvania, Philadelphia, PA 19104

Edited by Bruce S. McEwen, The Rockefeller University, New York, NY, and approved February 6, 2013 (received for review January 4, 2013)

Maternal stress is a key risk factor for neurodevelopmental disor-ders, including schizophrenia and autism, which often exhibit a sexbias in rates of presentation, age of onset, and symptom severity.The placenta is an endocrine tissue that functions as an importantmediator in responding to perturbations in the intrauterine en-vironment and is accessible for diagnostic purposes, potentially pro-viding biomarkers predictive of disease. Therefore, we have used agenome-wide array approach to screen placental expression acrosspregnancy for gene candidates that are sex-biased and stress-responsive in mice and translate to human tissue. We identifedO-linked-N-acetylglucosamine (O-GlcNAc) transferase (OGT), an X-linked gene important in regulating proteins involved in chroma-tin remodeling, as fitting these criteria. Levels of both OGT and itsbiochemical mark, O-GlcNAcylation, were significantly lower inmales and further reduced by prenatal stress. Examination of humanplacental tissue found similar patterns related to X chromosomedosage. As a demonstration of the importance of placental OGT inneurodevelopment, we found that hypothalamic gene expressionand the broad epigeneticmicroRNAenvironment in theneonatal brainof placental-specific hemizygous OGT mice was substantially altered.These studies identified OGT as a promising placental biomarker ofmaternal stress exposure that may relate to sex-biased outcomes inneurodevelopment.

O-glycosylation | extra-embryonic tissue | neuropsychiatric disorders

Maternal stress has been identified as a key risk factor forneurodevelopmental disorders, including schizophrenia andautism, which often exhibit a sex bias in rates of presentation, ageof onset, and symptom severity (1–4). Specifically, males exposedto stress during the first trimester have an increased risk forschizophrenia, suggesting early pregnancy may be a sensitivewindow of developmental vulnerability (5). Such epidemiologicalfindings also support a sex specificity of effects during a period ofrapid fetal and placental development. Further, sex differencesin gene expression in the placenta may represent unique modesof increased disease risk or resilience (4, 6).Alterations in prenatal programming associated with neuro-

developmental disorders likely involve complex interactions be-tween the maternal environment, the placenta, and factors of thedeveloping fetus, including sex (4, 6). In eutharian mammals, in-cluding humans and mice, the placenta serves to mediate commu-nication between the maternal and fetal compartments, deliveringnutrients and oxygen and protecting the developing fetus fromenvironmental insults (7). Although derived from both maternaland fetal contributions, the majority of the placenta develops outof the trophoblastic lineage of fetal origin, yielding a predominantlyXX or XY placenta, from which sex differences in response tootherwise similar intrauterine perturbations may emerge (8, 9).In our mouse model of early prenatal stress (EPS), we have

previously established that male offspring exposed to maternalstress during early gestation exhibit endophenotypes associatedwith neurodevelopmental disease, including stress dysregulationand cognitive deficits that persist into the second generation,confirming early pregnancy as a point of increased susceptibilityto the effects of maternal stress experience (5, 10–12). Dynamic

growth and programming of the placenta occurs during this earlyperiod, and therefore alterations in its function resulting frommaternal stress exposure could ultimately affect neurodevelopment(13–16). Therefore, we have used a genomic and proteomics ap-proach to identify potential placental biomarkers predictive ofprenatal stress exposure. Our proposed criteria for biomarkersassumed that candidate genes would (i) show a significant sexdifference in expression, (ii) be significantly altered in our EPSmodel, and (iii) be similarly regulated in human placental tissue.These criteria were established based upon the hypothesis that asex bias in disease risk stems from basal sex differences in abilitiesto adapt or respond to a perturbed environment.

ResultsEffects of Offspring Sex and EPS on Placental Gene Expression. Toidentify placental genes with sex differences in patterns of expres-sion consistent across gestation [embryonic day (E) 12.5, E15.5,and E18.5], we used a genome-wide microarray analysis. Thesegestational time points were selected because they reflect maturingstages of fetal development, representing likely important changesin placental function and interactions with the maternal environ-ment (17). Therefore, we hypothesized that identification of geneswith sex differences in expression consistent across this periodwould indicate important candidates potentially responsible forsex differences in programming outcomes in response to a per-turbed maternal environment. Differential expression analyses [falsediscovery rate (FDR) < 0.05] revealed dynamic developmentalchanges in gene expression patterns, with 2,658 placental geneschanging expression from E12.5 to E18.5. Surprisingly, sex differ-ences were only found for 11 X- and Y-linked genes, of which 4 areX and Y paralogs (Fig. 1 A and B and Table S1). Eight of thesegenes exhibited significant sex differences in patterns of expres-sion across pregnancy (Table S1). As expected, X-linked geneswere higher in females, and Y-linked genes were higher in males.To determine any effects of EPS on sexually dimorphic genes,

messenger RNA (mRNA) was compared between male andfemale control or EPS placentas (Fig. 1C). One gene, O-linked-N-acetylglucosamine transferase (OGT), an X-linked gene with lowerbaseline expression in male placentas, also had a reduction of ex-pression with the administration of EPS (Fig. 1D and Dataset S1). Itshould be noted that EPS had no effect on litter size or sex ratiosin this study (Dataset S1), as was previously reported (10).

Characterization of Placental OGT Protein, Enzymatic Activity, andChromatin State at the Ogt Locus. To determine any chromatin reg-ulation associated with the locus, we measured histone H3 tri-methyl Lys4 (H3K4me3), a permissive chromatin mark, at thepromoter region (Fig. 1E). Consistent with the basal sex differences

Author contributions: C.L.H. and T.L.B. designed research; C.L.H., C.P.M., and D.B.F. per-formed research; C.L.H. and C.P.M. analyzed data; and C.L.H. and T.L.B. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed. E-mail: tbale@vet.upenn.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1300065110/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1300065110 PNAS | March 26, 2013 | vol. 110 | no. 13 | 5169–5174

NEU

ROSC

IENCE

Dow

nloa

ded

by g

uest

on

June

17,

202

1

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1300065110/-/DCSupplemental/pnas.201300065SI.pdf?targetid=nameddest=ST1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1300065110/-/DCSupplemental/pnas.201300065SI.pdf?targetid=nameddest=ST1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1300065110/-/DCSupplemental/sd01.xlsxhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1300065110/-/DCSupplemental/sd01.xlsxmailto:tbale@vet.upenn.eduhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1300065110/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1300065110/-/DCSupplementalwww.pnas.org/cgi/doi/10.1073/pnas.1300065110

A BF12vsM12 F15vsM15

F18vsM18

0

1

0

2

0

0

8

E12vsE15 E15vsE18

E12vsE18

2658

646

2945

1319

5545

689

3801

C D

E F

Rel

ativ

e E

xpre

ssio

n

0.0

0.5

1.0

1.5

2.0

C EPS C EPS

*

#

F M

0 0.5 1.0 1.5 2.0Relative Expression

F/CF/EPSM/CM/EPS

XistKdm5c

Eif2s3xPramel3

Taf1

Rpsa

Ogt

Ddx3yUty

Eif2s3y

Kdm5d *

*

*

*

*

*

*

*

*

#

% o

f Inp

ut D

NA

0.0

0.5

1.0

1.5

2.0

C EPS C EPS

*

F M

Fig. 1. Identification of OGT as a potential biomarker of maternal stress. (A) Venn diagram of developmental differential expression analyses of mouseplacental genes via Affymetrix GeneChip Mouse Gene 1.0 ST microarray. The labels of each circle refer to specific embryonic time point comparisons rep-resented: E12vsE15 are the comparisons between E12.5 (n = 6) and E15.5 (n = 6), E15vsE18 are the comparisons between E15.5 and E18.5 (n = 6), and E12vsE18are the comparisons between E12.5 and E18.5. The numbers within the diagram represent the number of genes characterized as having differential expressionbetween these groups, independent of sex, using a false discovery rate of 0.05. (B) Venn diagram of sex differences in mRNA expression levels analyses ofmouse placental genes via microarray. Circle labels refer to the sex comparisons made during each developmental time point. The numbers are derived as inA. (C) Genes identified as having sex differences in expression examined in our EPS model. C, control; F, female; M, male (n = 7); EPS, early prenatal stress (n =8). Data for Xistwere normalized to the female control levels; data for all other traits were normalized to male control levels. Bars are the maximum likelihoodestimate for each group ± the 95% confidence interval for that estimate. Symbols (* for sex and # for EPS) indicate a main effect with a confidence intervalthat does not bound zero as determined by the linear model (y ∼ Sex + EPS + Sex*EPS). (D) OGT was identified as fitting our biomarker criteria for havingboth sex differences in expression and responding to EPS. Symbols (* for sex and # for EPS) indicate a main effect with a confidence interval that does notbound zero as determined by the linear model (y ∼ Sex + EPS + Sex*EPS). Normalization is as in C. (E) Diagram of the promoter region of OGT used for ChIPanalysis for presence at H3K4me3. (F) ChIP analysis of the OGT promoter demonstrating the expression changes in OGT are associated with the presence orabsence of the transcriptional activator mark, H3K4me3. Data are representative of the amount of DNA (normalized to the amount of input ornonimmunoprecipitated DNA) from the promoter region of OGT associated with the H3K4me3 chromatin mark. Bars are the maximum likelihood estimatefor each group ± the 95% confidence interval for that estimate. Asterisk indicates a measurable difference between female control and all other groups asdetermined by nonoverlapping confidence intervals for these estimates.

5170 | www.pnas.org/cgi/doi/10.1073/pnas.1300065110 Howerton et al.

Dow

nloa

ded

by g

uest

on

June

17,

202

1

www.pnas.org/cgi/doi/10.1073/pnas.1300065110

and EPS pattern of expression found for OGT, we detected a de-creased association of this locus with H3K4me3 in both male andin EPS placentas by chromatin immunoprecipitation (ChIP) (Fig.1F and Dataset S1).Changes in OGT protein levels corresponded with those found

for mRNA, with less protein in males compared with females,and less protein in male EPS placentas compared with controlmales (Fig. 2 A and B and Dataset S1). We compared total levelsof O-GlcNAc modified proteins as a biochemical readout of OGTenzymatic activity. We observed robust decreases of this mark inmale placentas compared with females (Fig. 2C and Dataset S1)but no overall difference in O-GlcNAcylation between male con-trol or EPS placentas (Dataset S1). Despite no overall difference,two bands at 28 and 37 kDa, which were visibly different betweenmale control and EPS placentas, were excised for proteomicanalysis (Fig. 2D). These proteins were identified as annexin A1(ANXA1) and peroxiredoxin 1 (PRDX1) (25 and 21, respectively,unique peptide sequences identified with >95% probability forthese proteins; Fig. 2 E and F and Table S2). Total protein levels ofboth ANXA1 and PRDX1 were not affected by EPS (Fig. S1 A andB and Dataset S1). There were no global sex differences in levelsof total serine or threonine phosphorylation, demonstrating thatall posttranslational modifications were not affected in the samemanner as O-GlcNAcylation (Fig. S2 A and B and Dataset S1).

Effect of X Chromosome Complement on Human Placental OGT mRNAand Protein. To determine the translational potential of our find-ings, OGT levels were evaluated in human term placenta. Biopsieswere obtained from male placentas, enriched for fetal (XY) ormaternal (XX) contributions to assess X-chromosomal comple-ment effects. OGT was highly expressed in human placenta fromboth maternal and fetal contributions. Similar to our findings in

mice, gene expression in XY samples was measurably lower forOGT compared with XX samples (Fig. 2G and Dataset S1).Biochemical analysis of O-GlcNAc modified proteins followedthis same pattern (Fig. 2H and Dataset S1).

Reduced Placental OGT Results in Broad Neurodevelopmental Changes.To determine the potential programming effects that reducedplacental OGT would impose on the developing brain, we used aplacental-specific Cre-recombinase–expressing mouse to condi-tionally target the Ogt gene (18). Using this genetic model, femalemice had reduced placental OGT expression but no differencein the liver (Dataset S1). We examined broad patterns of geneexpression in the neonatal hypothalamus, because we weremost interested in changes that would potentially affect neu-roendocrine systems. In addition, we compared neonatal whole-brain microRNA patterns as a marker of the potential differencesin the epigenetic environment in the developing brain. BecauseOGT is X-linked, wild-type and hemizygous females were com-pared because our goal was to examine a model of controlledreduced OGT, and not a knockout of this gene, which would bethe result in males. We detected robust changes in hypothalamicgene regulation, with 370 genes showing significant differences(FDR < 0.05; Fig. 3A and Dataset S2). Further, functional anno-tation clustering of these affected genes revealed a significantenrichment for pathways involved in energy utilization, proteinregulation, and synapse formation, processes important in neuro-development (Fig. 3A) (19, 20). Similar to the extensive gene ex-pression changes we detected in the hypothalamus, we also foundrobust differences in the brain microRNA environment, wherehemizygous placental OGT expression shifted the pattern of 250of the most abundant brain microRNAs to be distinct from that ofwild-type females by hierarchical clustering analyses (Fig. 3B).

A B

C D

E

F

G H

Fig. 2. Biochemical assessment of OGT and O-GlcNAcylationin mouse and human placentas. (A) Representative Westernblot images of OGT levels from male and female mouse pla-centas. Histogram is the maximum likelihood estimate for thenormalized optical densities of each group (n = 7) ± the 95%confidence interval for that estimate. Asterisk indicatesmeasurable difference between groups as determined bynonoverlapping confidence intervals for the estimates. (B)Representative Western blot images of OGT levels fromcontrol (n = 7) and EPS (n = 8) male mouse placentas. Histo-gram was derived and annotated as in A. (C ) RepresentativeWestern blot images of total O-GlcNAcylated proteins frommale and female mouse placentas. Histogram was derived andannotated as in A (n = 7). (D) Representative Western blotimages of total O-GlcNAcylated proteins from male control (n =7) and EPS (n = 8) placentas. The image is annotated to high-light the bands visibly identified with differential O-GlcNA-cylation between treatment groups excised for proteomicanalyses. (E ) Amino acid sequence coverage of LC-MS/MSwhere the band at 28 kDA was identified as peroxiredoxin-1(PRDX1) and the band at 37 kDA was identified as annexin A1(ANXA1) via LC-MS/MS with greater than 99.9% certainty. (F)Representative spectrum image of peptide fragments used toidentify both ANXA1 and PRDX1. (G) Expression of OGT by RT-PCR in human placental tissue associated with male births. Datawere normalized to XY levels. Bars are the maximum likeli-hood estimate for each group (n = 4) ± the 95% confidenceinterval for that estimate. XX, maternal; XY, fetal. (H) Rep-resentative Western blot images of total O-GlcNAcylatedproteins from both XX and XY placental contributions fromhuman placentas. Histogram was derived as in A (n = 4). As-terisk indicates a measurable difference between groups withdifferent X-chromosome complement as determined by non-overlapping confidence intervals for each estimate.

Howerton et al. PNAS | March 26, 2013 | vol. 110 | no. 13 | 5171

NEU

ROSC

IENCE

Dow

nloa

ded

by g

uest

on

June

17,

202

1

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1300065110/-/DCSupplemental/sd01.xlsxhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1300065110/-/DCSupplemental/sd01.xlsxhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1300065110/-/DCSupplemental/sd01.xlsxhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1300065110/-/DCSupplemental/sd01.xlsxhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1300065110/-/DCSupplemental/pnas.201300065SI.pdf?targetid=nameddest=ST2http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1300065110/-/DCSupplemental/pnas.201300065SI.pdf?targetid=nameddest=SF1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1300065110/-/DCSupplemental/pnas.201300065SI.pdf?targetid=nameddest=SF1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1300065110/-/DCSupplemental/sd01.xlsxhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1300065110/-/DCSupplemental/pnas.201300065SI.pdf?targetid=nameddest=SF2http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1300065110/-/DCSupplemental/sd01.xlsxhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1300065110/-/DCSupplemental/sd01.xlsxhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1300065110/-/DCSupplemental/sd01.xlsxhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1300065110/-/DCSupplemental/sd01.xlsxhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1300065110/-/DCSupplemental/sd02.xlsx

DiscussionAn increased risk for neurodevelopmental disorders such asschizophrenia has been associated with fetal antecedents in-cluding prenatal stress. Because such diseases often present witha large sex bias in rates of diagnosis, age of onset, and symptomseverity, we used a genomics approach to identify potentialplacental candidate genes predictive of prenatal stress exposurefrom male and female placental tissue taken at E12.5, E15.5, andE18.5 (1–4). From this screen, OGT was identified as an importantplacental effector of programmatic changes in neurodevelopment,providing a potential mechanism for neurodevelopmental changesresulting from maternal stress in early pregnancy (EPS). The cri-teria used in this assessment included that the candidate gene must(i) show a significant sex difference in expression across preg-nancy, (ii) be significantly altered in our EPS model, and (iii) besimilarly regulated in human placental tissue. The hypothesis inthese studies centered on a potential threshold effect wherebybasal sex differences in placental gene expression may provide asex bias for risk or resiliency in response to prenatal insults.OGT is a key cellular regulator through the unique post-

translational modification it places on serine and threonine res-idues of both nuclear and cytosolic proteins. More specifically,OGT plays an important role in programmatic chromatin re-modeling by O-glycosylation (O-GlcNAcylation) of protein tar-gets, including RNA polymerase II, histone deacetylases, andhistone 2B (21). Because O-GlcNAcylation competes with ser-ine/threonine phosphorylation, it serves a critical function in theregulation of enzymatic activity important in somatic cell func-tion and embryo viability (22–25). Additionally, OGT has beencharacterized as a cellular nutrient sensor (26) and therefore mayplay an important role in the protective effects of the placenta onthe developing brain from insults such as maternal food depri-vation (7). We hypothesized that levels of this important enzymebelow a critical point following maternal stress could significantlyaffect specific aspects of placental function. Therefore, becausemale placentas showed a lower baseline expression of OGT, astress-mediated further reduction may specifically place malefetuses at a disadvantage in being able to adapt to a changingenvironment, and at an increased risk for long-term effects on

neurodevelopment. Additionally, OGT has been characterizedas a cellular nutrient sensor (26).Levels of OGT mRNA, protein, and O-GlcNAc were all re-

lated to X-chromosome complement, suggesting this gene likelyescapes X inactivation in the placenta. These levels were allfurther reduced in our mouse model of EPS. Factors modulatingOGT expression and translation have not been well character-ized. However, biallelic expression in female embryonic stemcells at varying stages of differentiation has been reported (27,28). As further evidence for regulation of OGT expression at atranscriptional level, ChIP was conducted at the Ogt locus withH3K4me3, a permissive chromatin mark indicative of transcrip-tional activation. A pattern of association with this activationalmark was found similar to that detected for OGT mRNA andprotein levels, where there was a reduction for males and EPSgroups compared with control females. Interestingly, the his-tone demethylases, lysine-specific demethylase 5C (KDM5C)and lysine-specific demethylase 5D (KDM5D), which demethylateH3K4me3, were genes also identified as having a sex-dependentexpression pattern in our placental array. Taken together, thesedata support additional modes of sex-dependent epigenetic reg-ulation whereby a potential imbalance in chromatin regulationmay affect placental function.In addition to OGT identification in this genomic screen as an

important candidate placental biomarker, additional genes ofinterest were also noted. Overall, there were dynamic changes ingene expression patterns and within unique gene sets over thethree gestational time points examined. Despite the dynamicnature of this tissue, only a small subset of genes demonstratedsex differences in patterns of expression, all of which were sexchromosome-linked: Ogt; X-inactive specific transcript (Xist);Kdm5c; eukaryotic translation initiation factor 2, subunit 3,structural gene X-linked (Eif2s3x); preferentially expressed anti-gen in melanoma-like 3 (Pramel3); TATA box binding protein-associated factor 1 (Taf1); ribosomal protein SA (Rpsa); DEADbox polypeptide 3, Y-linked (Ddx3y); ubiquitously transcribedtetratricopeptide repeat containing, Y-linked (Uty); eukaryotictranslation initiation factor 2, subunit 3, structural gene Y-linked(Eif2s3y); and Kdm5d. These genes have functions involved inprotein translation, histone methylation, and protein glycosylation

A B

NADH ProcessingRibosomal FunctioningElectron Transport Chain

Mitochondrial Structure

Nuclear Envelope

Functional groups downregulatedwith reduced placental Ogt

UbiquitinationReceptor SignallingSynaptic Plasticity

WD40 Domain Proteins

Proteolysis

Functional groups upregulatedwith reduced placental Ogt

F P.Cre -F P.Cre + F P.Cre -

F P.Cre +

Fig. 3. Placental specific reduction of OGT broadly affectsearly neurodevelopment. (A) Heat map of PN2 hypothalamicgenes with significantly different levels of expression be-tween XOgt/ XWT females, with or without placental Creexpression [P.Cre+ (n = 4) or P.Cre– (n = 5), respectively],resulting in a 50% reduction in OGT expression in the P.Cre+females (Fig. S3). Genes that were either significantly up-regulated (n = 218) or significantly down-regulated (n = 161)with reduced placental OGT, as determined by a FDR of 0.01,were grouped into functional annotation categories using theDAVID functional annotation clustering tool. (B) Heat map ofwhole-brain microRNA expression levels from the same ani-mals in A. Hierarchical clustering discriminates all but onesample between the two groups [P.Cre+/− (n = 5) vs. P.Cre−/−

(n = 4)], suggesting broad epigenetic changes in the PN2 brain.

5172 | www.pnas.org/cgi/doi/10.1073/pnas.1300065110 Howerton et al.

Dow

nloa

ded

by g

uest

on

June

17,

202

1

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1300065110/-/DCSupplemental/pnas.201300065SI.pdf?targetid=nameddest=SF3www.pnas.org/cgi/doi/10.1073/pnas.1300065110

(Table S1). Although these genes showed significant sex differ-ences, no further effect of EPS was detected. However, the sexdifferences in their patterns of expression suggest that they mayplay an important role in sex differences in placental function thatcould also contribute to a sex bias in neurodevelopmental disease.Further, the surprisingly limited number of genes exhibiting sexdifferences in expression in the placenta across pregnancy usingour strict statistical requirements highlights the unique control ofgene expression in the placenta and the potential impact pertur-bations during gestation may have on sex biases in programmingoutcomes in the developing fetus.In the biochemical assessment of OGT activity, O-GlcNAc

levels were significantly reduced in male placentas, fitting withtheir decreased OGT mRNA and protein. In response to EPS,we noted two distinct bands in male placental samples that wereapparently reduced. These proteins were identified as ANXA1and PRDX1 by proteomics analyses. Both proteins are knowntargets of OGT; however, the exact site of O-glycosylation hasnot been determined (29). ANXA1 is an important modulator ofcellular inflammatory signaling, and PRDX1 is an intracellularperoxidase that functions as a chaperone with clients includingnuclear factor-kappa β (30–34). These are two specific examplesof mechanisms by which EPS may produce sex differences in pla-cental function that could affect brain development trajectories.To determine the translational potential of our findings, OGT

levels and biochemistry were evaluated in human term placentaltissue. Biopsies obtained from male placentas enriched for fetal(XY) or maternal (XX) contributions to assess X-chromosomalcomplement found that OGT was highly expressed and, similarto our findings in mice, XY samples were significantly lower forOGT compared with XX samples. Biochemical analysis of O-GlcNAcylated proteins followed this same pattern (Fig. 2H).Human placental tissue was obtained from deidentified cesarean-section deliveries, and thus the pregnancy experience and exposureto various stressors or other fetal antecedents could not bedetermined. As such, to avoid likely confounding of these un-controllable variables, only X-chromosomal dosage effects onOGT expression were evaluated (i.e., frommale births). Together,these data support OGT as meeting all three of our criteria asa predictive placental biomarker.To examine a more direct link between reduced placental OGT

and potential programming changes in the developing brain, weused a placental-specific Cre-recombinase–expressing mouse toconditionally target the Ogt gene. We examined broad patternsof gene expression in the neonatal hypothalamus, because wewere interested in programming changes that would potentiallyaffect neuroendocrine systems based on our EPS model, byAffymetrix microarray (12, 35). In addition, we compared neo-natal whole-brain microRNA patterns by ABI Taqman array asa marker of the potential differences in the epigenetic environ-ment in the developing brain, and based on our previous findingsfor dynamic microRNA changes in our EPS model (35). BecauseOGT is X-linked, wild-type and hemizygous females were com-pared to examine a model of controlled reduced OGT, and not aknockout of this gene, which would be the result in males. Im-portantly, gene expression changes as a result of reduced pla-cental OGT in this context are representative of the impact thisbiomarker has on neurodevelopment irrespective of chromo-somal sex and were not directly compared with the effects of EPSon early brain development. We detected profound changes inhypothalamic gene regulation, with over 375 genes showing sig-nificant differences in expression patterns in the hypothalamusfrom placental hemizygous Ogt mice compared with controls.Further, functional annotation clustering of these affected genesrevealed a significant enrichment for pathways involved in energyutilization, protein regulation, and synapse formation, all pro-cesses important in neurodevelopment (19, 20). Similar to theextensive gene expression changes we detected in the hypothal-amus, we also found robust differences in the brain microRNAenvironment, where hemizygous placental OGT expression shiftedthe pattern of 250 of the most abundant brain microRNAs to be

distinct from that of control animals by hierarchical clusteringanalyses. These results support our previous studies showing similarbroad changes produced by EPS in the postnatal brain microRNAenvironment indicative of epigenetic programming involvement inshaping plasticity during neonatal brain development (35). Thesedata substantiate the importance of regulation of placental OGTlevels and provide a potential link with profound changes in genetranscription in neurodevelopment with a candidate biomarker.The endocrine placenta is poised to be a key mediating tissue

in responding to a dynamic and changing maternal environment.Sex differences in how this tissue responds to the same maternalmilieu likely contribute to the altered susceptibility of male andfemale fetuses to long-term programming outcomes (4). Ourstudies identified OGT, an intracellular glycotransferase impor-tant in regulating key chromatin programming events, as a po-tential placental biomarker that was sex-biased in its expression,responsive to EPS, and similar in X-linked expression pattern inhuman tissue. Because the placenta is readily accessible, thesestudies have high translational potential and may be useful inpredicting gestational stress exposure.

MethodsAnimals. Male C57BL/6J and female 129S1/SvImJ mice were obtained fromJackson Laboratories and subsequently used as breeding stock to produceC57BL/6J:129S1/SvImJ hybrids (F1 hybrids). F1 hybrid breeding pairs (n = 33)were checked daily at 7:00 AM for copulation plugs. Noon on the day thatthe plug was observed was considered to be embryonic day 0.5 (E0.5). F1hybrids were used for the placental microarray and the prenatal stressexperiments. For the placental-specific reduction of OGT, double heterozy-gous [B6.129-Ogttm1Gwh/J (XOgt/ XWT); B6-CYP19-Cre (placental-Cre recombi-nase heterozygote, P.Cre+/−)] females (n = 17) were bred to hemizygousB6.129-Ogttm1Gwh/J (XOgt/ Y)/ heterozygous (P.Cre+/−) males (n = 14),resulting in offspring representing all potential genotypes. To identify theeffects of a placental-specific reduction of OGT, female XOgt/ XWT P.Cre+/−

(P.Cre+/−, n = 10) were compared with XOgt/ XWT P.Cre−/− (P.Cre+/−, n = 10)for placental specificity validation and brain analyses. All animal protocolswere reviewed and approved by the Institutional Animal Care and Use Com-mittee of the University of Pennsylvania.

Early Prenatal Stress. Administration of chronic variable stress was performedas previously described (12). Dams were randomly assigned to treatmentgroups to receive stress during days 1–7 of gestation (EPS; n = 8) or to acontrol (n = 7) nonstressed group. Pregnant mice assigned to the EPS groupexperienced each of the following stressors on a different day of the EPSperiod: 60 min (beginning at 1:00 PM) of fox odor exposure (1:5,000 2,4,5-trimethylthiazole; Acros Organics), 15 min of restraint (beginning at 1:00 PM)in a modified 50-mL conical tube, 36 h of constant light, novel noise (WhiteNoise/Nature Sound-Sleep Machine; Brookstone) overnight, three cagechanges (at ∼9:00 AM, 12:00 PM, and 3:00 PM) throughout the light cycle,saturated bedding (700 mL, 23 °C water) overnight, and novel object (eightmarbles of similar shape and color) exposure overnight. These stressors wereselected to be nonhabituating and did not induce pain or directly influencematernal food intake, weight gain, or behavior (10).

Mouse Tissue Dissection. For the placental microarray experiment mouseplacentas and embryonic somatic tissue were dissected at E12.5, E15.5, andE18.5 (n =6 at each gestational time point). DNA was extracted from theembryonic somatic tissue, and the sex of the embryo was determined usingmethods described elsewhere (36). One male and one female placenta fromeach litter were used for microarray analysis; these placentas were quarteredand put in 500 μL of RNAlater and stored at −80C until processed as below.For the prenatal stress experiment, E12.5 male and female placentas fromcontrol (n = 7) and EPS (n = 8) litters were used for gene expression, ChIP,and protein analyses; these placentas were bisected via transverse sectioningand one half was placed in 500 μL of RNAlater and stored at −80C (for geneexpression analyses) and one half was flash-frozen in liquid nitrogen andstored at −80C (for ChIP and Western blot analyses).

For the placental-specific reduction of OGT experiment, two cohorts ofanimals were used: (i) an embryonic cohort (n = 4 P.Cre+/− and P.Cre−/−) tovalidate the specificity of OGT reduction to the placenta and (ii) a cohort (n = 4P.Cre+/−; n = 5 P.Cre−/−) killed on postnatal day 2 (PN2) for brain analyses. Forthe validation, placentas and embryonic somatic tissue were dissected at E12.5.DNA was extracted as above; placentas and liver tissue were flash-frozen in

Howerton et al. PNAS | March 26, 2013 | vol. 110 | no. 13 | 5173

NEU

ROSC

IENCE

Dow

nloa

ded

by g

uest

on

June

17,

202

1

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1300065110/-/DCSupplemental/pnas.201300065SI.pdf?targetid=nameddest=ST1

liquid nitrogen and stored at −80C. For the brain analyses, pups were killedat PN2 and brains were dissected and flash-frozen in liquid nitrogen andstored at −80C. Tissue punches (300 μm) of the hypothalamus [∼3.0–3.3 mmposterior in the anterior–posterior plane and 1.5 mm ventral in the dorsal–ventral plane, structures corresponding to the paraventricular area in a de-velopmental mouse brain atlas (37)] were collected using a 1.0-mm circularpunch (Ted Pella) and then used for microarray analysis; the remaining braintissue was used for microRNA analysis.

Human Tissue Samples. Placental tissues associated with deidentified malefetuses (n = 4) were quartered by cutting in the transverse section throughthe location of the emergence of the umbilical cord in two directions. Bi-opsies (∼5 mg) were obtained from both the fetal and maternal con-tributions to the placenta from three locations, 5 cm from the umbilical cord.Biopsies were flash-frozen in liquid nitrogen and stored at −80 C until fur-ther analysis. Human tissue collection protocols were reviewed and ap-proved by the Institutional Board of the University of Pennsylvania.

Brain microRNA Environment. Total RNA was extracted from brains [P.Cre+/−

(n = 4) and P.Cre−/− (n = 5)] using TRIzol reagent (Invitrogen). One micro-gram of total RNA was reverse-transcribed to cDNA using Megaplex RT poolA primers and Multiscribe reverse transcriptase (Applied Biosystems). Ex-pression levels of 245 microRNAs were determined using the Taqman ArrayMicroRNA card A Array (Applied Biosystems). Analysis was performed usingthe comparative cycle threshold (Ct) method. For each sample, the averageof the Ct values of small nucleolar RNA (sno)135 and sno202 was used as anendogenous loading control. Expression levels of each sample were normalized

to the average expression level of P.Cre−/− females. Uninformed hierarchicalclustering using Pearson correlations was used to discriminate differentialmicroRNA expression between individuals (42).

Statistical Analyses. Analyses were performed using R (version 2.14.2) and thepackages arm, (43), bbmle (44), and limma (39, 45) to fit the gene expressionand optical density data to linear models, and estimates for main and inter-action effects were determined from these models. For all analyses, maximumlikelihood estimates were calculated and 95% confidence intervals wereconstructed for regression coefficients as the reported statistical values.We used this statistical approach over the more traditional null hypothesissignificance testing to provide the magnitude for any observed experi-mental effect, to provide a measure of dispersion around this effect sizeto allow assessment of the predictive value of our models, and to relateany statistical conclusions to the experimental hypothesis (46). A confi-dence interval for an estimate that did not bound zero was considered tobe significant; comparisons among experimental groups were made usingconfidence interval evaluation, and those that did not overlap were con-sidered to be significantly different. Database for Annotation, Visualizationand Integrated Discovery (DAVID) functional annotation clustering (47)was used for the PN2 hypothalamic analysis, and MeV v4.8 was used forthe PN2 microRNA analysis.

ACKNOWLEDGMENTS. We thank C. Taylor for technical assistance andanimal care and G. Leone at The Ohio State University for the generousdonation of Cyp19-Cre mice. This work is supported by National Institutes ofHealth Grants MH099910, MH091258, and MH087597.

1. (2012) Prevalence of autism spectrum disorders–Autism and Developmental DisabilitiesMonitoring Network, 14 sites, United States, 2008.MMWR CDC Surveill Summ 61(3):1–19.

2. Ober C, Loisel DA, Gilad Y (2008) Sex-specific genetic architecture of human disease.Nat Rev Genet 9(12):911–922.

3. Bale TL, et al. (2010) Early life programming and neurodevelopmental disorders. BiolPsychiatry 68(4):314–319.

4. Howerton CL, Bale TL (2012) Prenatal programing: At the intersection of maternalstress and immune activation. Horm Behav 62(3):237–242.

5. Khashan AS, et al. (2008) Higher risk of offspring schizophrenia following antenatalmaternal exposure to severe adverse life events. Arch Gen Psychiatry 65(2):146–152.

6. Bale TL (2011) Sex differences in prenatal epigenetic programming of stress pathways.Stress 14(4):348–356.

7. Broad KD, Keverne EB (2011) Placental protection of the fetal brain during short-termfood deprivation. Proc Natl Acad Sci USA 108(37):15237–15241.

8. Rossant J, Cross JC (2001) Placental development: Lessons from mouse mutants. NatRev Genet 2(7):538–548.

9. Rosenfeld CS (2012) Effects of maternal diet and exposure to bisphenol A on sexuallydimorphic responses in conceptuses and offspring. Reprod Domest Anim 47(Suppl 4):23–30.

10. Mueller BR, Bale TL (2006) Impact of prenatal stress on long term body weight isdependent on timing and maternal sensitivity. Physiol Behav 88(4-5):605–614.

11. Mueller BR, Bale TL (2007) Early prenatal stress impact on coping strategies andlearning performance is sex dependent. Physiol Behav 91(1):55–65.

12. Mueller BR, Bale TL (2008) Sex-specific programming of offspring emotionality afterstress early in pregnancy. J Neurosci 28(36):9055–9065.

13. Marsit CJ, Maccani MA, Padbury JF, Lester BM (2012) Placental 11-beta hydroxysteroiddehydrogenase methylation is associated with newborn growth and a measure ofneurobehavioral outcome. PLoS ONE 7(3):e33794.

14. O’Donnell KJ, et al. (2012) Maternal prenatal anxiety and downregulation of pla-cental 11β-HSD2. Psychoneuroendocrinology 37(6):818–826.

15. Pena CJ, Monk C, Champagne FA (2012) Epigenetic effects of prenatal stress on 11beta-hydroxysteroid dehydrogenase-2 in the placenta and fetal brain. PLoS ONE 7(6).

16. Wyrwoll CS, Kerrigan D, Holmes MC, Seckl JR, Drake AJ (2012) Altered placental methyldonor transport in the dexamethasone programmed rat. Placenta 33(3):220–223.

17. Keverne EB (2010) Epigenetically regulated imprinted genes and foetal programming.Neurotox Res 18(3-4):386–392.

18. Wenzel PL, Leone G (2007) Expression of Cre recombinase in early diploid trophoblastcells of the mouse placenta. Genesis 45(3):129–134.

19. Schmidt-Kastner R, van Os J, Esquivel G, Steinbusch HWM, Rutten BPF (2012) Anenvironmental analysis of genes associated with schizophrenia: Hypoxia and vascularfactors as interacting elements in the neurodevelopmental model. Mol Psychiatry17(12):1194–1205.

20. Sun J, et al. (2010) Schizophrenia gene networks and pathways and their applicationsfor novel candidate gene selection. PLoS ONE 5(6):e11351.

21. Fujiki R, et al. (2011) GlcNAcylation of histone H2B facilitates its monoubiquitination.Nature 480(7378):557–560.

22. Shafi R, et al. (2000) The O-GlcNAc transferase gene resides on the X chromosome andis essential for embryonic stem cell viability and mouse ontogeny. Proc Natl Acad SciUSA 97(11):5735–5739.

23. O’Donnell N, Zachara NE, Hart GW, Marth JD (2004) Ogt-dependent X-chromosome-linked protein glycosylation is a requisite modification in somatic cell function andembryo viability. Mol Cell Biol 24(4):1680–1690.

24. Zachara NE, et al. (2004) Dynamic O-GlcNAc modification of nucleocytoplasmic proteins inresponse to stress. A survival responseofmammalian cells. J Biol Chem 279(29):30133–30142.

25. Comer FI, Hart GW (2000) O-Glycosylation of nuclear and cytosolic proteins. Dynamicinterplay between O-GlcNAc and O-phosphate. J Biol Chem 275(38):29179–29182.

26. Lazarus MB, Nam YS, Jiang JY, Sliz P, Walker S (2011) Structure of human O-GlcNActransferase and its complex with a peptide substrate. Nature 469(7331):564–567.

27. Lin H, et al. (2007) Dosage compensation in the mouse balances up-regulation andsilencing of X-linked genes. PLoS Biol 5(12):e326.

28. Mao J, et al. (2010) Contrasting effects of different maternal diets on sexually dimorphicgene expression in the murine placenta. Proc Natl Acad Sci USA 107(12):5557–5562.

29. Wang J, Torii M, Liu H, Hart GW, Hu Z-Z (2011) dbOGAP – An integrated bio-informatics resource for protein O-GlcNAcylation. BMC Bioinformatics 12(1):91.

30. Wang XM, He SH, Sun JM, Delcuve GP, Davie JR (2010) Selective association of per-oxiredoxin 1 with genomic DNA and COX-2 upstream promoter elements in estrogenreceptor negative breast cancer cells. Mol Biol Cell 21(17):2987–2995.

31. Neumann CA, Cao JX, Manevich Y (2009) Peroxiredoxin 1 and its role in cell signaling.Cell Cycle 8(24):4072–4078.

32. Chhipa RR, Lee KS, Onate S, Wu Y, Ip C (2009) Prx1 enhances androgen receptorfunction in prostate cancer cells by increasing receptor affinity to dihydrotestoster-one. Mol Cancer Res 7(9):1543–1552.

33. Flower RJ, Blackwell GJ (1976) The importance of phospholipase-A2 in prostaglandinbiosynthesis. Biochem Pharmacol 25(3):285–291.

34. Parente L, Solito E (2004) Annexin 1: more than an anti-phospholipase protein. In-flamm Res 53(4):125–132.

35. Morgan CP, Bale TL (2011) Early prenatal stress epigenetically programs dysmasculinizationin second-generation offspring via the paternal lineage. J Neurosci 31(33):11748–11755.

36. Clapcote SJ, Roder JC (2005) Simplex PCR assay for sex determination in mice. Bio-techniques 38(5):702, 706, 704, 706.

37. Paxinos G, Halliday H, Watson C, Koutcherov Y, Wang H (2007) Atlas of the De-veloping Mouse Brain at E17.5, P0, and P6 (Academic Press, London).

38. Gautier L, Cope L, Bolstad BM, Irizarry RA (2004) affy—Analysis of AffymetrixGeneChip data at the probe level. Bioinformatics 20(3):307–315.

39. Smyth GK (2005) Limma: Linear models for microarray data. Bioinformatics andComputational Biology Solutions Using R and Bioconductor, eds Gentleman R, Carey V,Dudoit S, Irizarry RA, Huber W (Springer, New York), pp 397–420.

40. Irizarry RA, et al. (2003) Summaries of Affymetrix GeneChip probe level data. NucleicAcids Res 31(4):e15.

41. Benjamini Y, Hochberg Y (1995) Controlling the false discovery rate: A practical andpowerful approach to multiple testing. JR Stat Soc 57(1):289–300.

42. Berry MPR, et al. (2010) An interferon-inducible neutrophil-driven blood transcrip-tional signature in human tuberculosis. Nature 466(7309):973–977.

43. Gelman A, Hill J (2006) Data Analysis Using Regression and Multilevel/HierarchicalModels (Cambridge Univ Press, Cambridge, UK).

44. Bolker B (2010) bbmle: Tools for general maximum likelihood estimation. (TheComprehensive R Archive Network, Vienna).

45. Smyth GK (2004) Linear models and empirical bayes methods for assessing differentialexpression in microarray experiments. Stat Appl Genet Mol Biol 3(1):e3.

46. Nakagawa S, Cuthill IC (2007) Effect size, confidence interval and statistical signifi-cance: A practical guide for biologists. Biol Rev Camb Philos Soc 82(4):591–605.

47. Sherman BT, et al. (2007) DAVID Knowledgebase: A gene-centered database in-tegrating heterogeneous gene annotation resources to facilitate high-throughputgene functional analysis. BMC Bioinformatics 8:426.

5174 | www.pnas.org/cgi/doi/10.1073/pnas.1300065110 Howerton et al.

Dow

nloa

ded

by g

uest

on

June

17,

202

1

www.pnas.org/cgi/doi/10.1073/pnas.1300065110