Prenatal Alcohol Exposure Causes Persistent Microglial Activation and Age- and Sex- Specific Effects on Cognition and Metabolic Outcomes in an Alzheimer’s Disease Mouse Model

Abstract

Background:

Prenatal alcohol exposure (PAE) causes behavioral deficits and increases risk of metabolic diseases. Alzheimer’s Disease (AD) is a neurodegenerative disease that has a higher risk in adults with metabolic diseases. Both present with persistent neuroinflammation.

Objectives:

We tested whether PAE exacerbates AD-related cognitive decline in a mouse model (3xTg-AD; presenilin/amyloid precursor protein/tau), and assessed associations among cognition, metabolic impairment, and microglial reactivity.

Methods:

Alcohol-exposed (ALC) pregnant 3xTg-AD mice received 3g/kg alcohol from embryonic day 8.5–17.5. We evaluated recognition memory and associative memory (fear conditioning) in 8–10 males and females per group at 3 months of age (3mo), 7mo, and 11mo, then assessed glucose tolerance, body composition, and hippocampal microglial activation at 12mo.

Results:

ALC females had higher body weights than controls from 5mo (p<0.0001). Controls showed improved recognition memory at 11mo compared with 3mo (p=0.007); this was not seen in ALC mice. Older animals froze more during fear conditioning than younger, and ALC mice were hyper-responsive to the fear-related cue (p=0.017). Fasting blood glucose was lower in ALC males and higher in ALC females than controls. Positive associations occurred between glucose and fear-related context (p=0.04) and adiposity and fear-related cue (p=0.0002) in ALC animals. Hippocampal microglial activation was higher in ALC than controls (p<0.0001); this trended to correlate with recognition memory.

Conclusions:

ALC animals showed age-related cognitive impairments that did not interact with AD risk but did correlate with metabolic dysfunction and somewhat with microglial activation. Thus, metabolic disorders may be a therapeutic target for people with FASDs.

Introduction

Prenatal alcohol exposure (PAE) is the leading preventable cause of neurodevelopmental disability. Self-reports indicate that 13.5% pregnant women in the United States consume alcohol and 5.2% report binge drinking (1). This aligns with an estimated rate of Fetal Alcohol Spectrum Disorders (FASDs) in the USA of 1–5% (2). Children with FASDs often experience cognitive impairment (3–5) and adults with a FASD diagnosis have an increased incidence of age-related diseases including obesity and type-2 diabetes (T2D; 6–10). The risk of obesity is higher in adolescent and adult females with FASDs whereas adult males have greater risk for T2D (7,10–13), suggesting an interaction among PAE, age, and sex.

Preclinical models of FASD recapitulate these cognitive and metabolic outcomes (14–19). It is unknown whether the metabolic dysfunctions are related to the cognitive and behavioral deficits; however, we recently identified significant associations among metabolic outcomes and behaviors in aged mice exposed to alcohol prenatally, such that the severity of an individual’s adiposity and glucose intolerance positively correlates with the magnitude of its learning deficits (19).

Mechanisms contributing to these behavioral deficits and metabolic dysfunctions are complex, however, inflammation is a likely contributor (14,20–22). Inflammation is a likely mechanistic driver in obesity and T2D (23), and PAE is associated with a systemic proinflammatory state and risk for obesity and T2D (6,21,24,25). PAE can also lead to persistent neuroinflammation and hyperresponsive microglia later in life. Rodents with PAE have persistently increased microglial activation and proinflammatory cytokine production through the first three postnatal months (26–28). Microglia are the resident innate immune cells of the brain; they mediate inflammatory responses to injury or insults and help return the brain to homeostasis once the injury or insult is resolved (29). The persistent neuroinflammation observed following PAE may be a contributing factor to the behavioral deficits (22,25,30–32) and functional blockade of inflammatory signals improves cognitive performance and reduces neuroinflammation (27,32,33).

Alzheimer’s Disease (AD) is a progressive neurodegenerative disease that affects ~10% of the US population over the age of 65 (34). AD is characterized by a progressive decline in learning and memory and pathological characteristics that include chronic neuroinflammation (35–39) and the presence of amyloid plaques and hyperphosphorylated tau tangles (40,41). Genetic factors play a significant role in AD risk (42), and their influence is further modified by non-genetic factors, including chronic alcohol consumption, obesity, and diabetes (43–46).

There are multiple mouse models of AD; one commonly used incorporates a human mutation in presenilin-1 with overexpression of the amyloid beta precursor protein and a hyperphosphorylated tau (3xTg-AD; (47)). Pathological changes emerge in these mice around 7 months of age (mo) including neuroinflammation (e.g. microglial activation), amyloid plaques, and tau tangles (48–51). Cognitive deficits emerge around 4–6mo (48,52), with different domains showing effects at different ages. Deficits in associative learning (the ability to associate two or more separate stimuli) and working memory (the ability to retain and work with information over short time periods) are seen at 6–8mo, whereas deficits in recognition memory (the ability to remember a previously encountered stimulus or situation) deficits emerge around 9–11mo (50). Metabolic changes are sex-dependent in this model; adult females show increased adiposity and poorer glucose tolerance when consuming a control diet, whereas this metabolic phenotype is only apparent in males that consume a high fat diet (53).

Metabolic Syndrome (MetS) describes a metabolic dysfunction that includes T2D, obesity, and hypertension; MetS worsens with aging (54) and contributes to AD risk (55–62). Animal models show that MetS emerges in adult animals that experience PAE, with greater insulin resistance, adiposity, and hypertension compared to age-matched controls (14,15,63–66). Clinical studies also report greater obesity, elevated body mass indices (BMIs), and hypertension in those with PAE compared with normotypic controls (10,12,67–69). Importantly, MetS and inflammation are causally linked in a feed-forward loop; chronic inflammation drives MetS and its pathological consequences that further promote inflammatory responses (23).

Given that PAE promotes inflammation and MetS, and that these are mechanistically entwined with AD, we hypothesized that PAE exacerbates the cognitive decline that is a feature of AD, and this is associated with metabolic impairment and neuroinflammation. We test this hypothesis using the 3xTg-AD mouse model and assess behavior outcomes following PAE at three different adult ages including prior to the emergence of pathological changes (3mo), around the emergence of the pathological changes (7mo), and a later age where cognitive impairment is more advanced (11mo) (48,70). We also assess the impact of PAE on metabolic outcomes and microglial activation in these same animals at 12mo, and evaluate associations among behavior, metabolic, and microglial activation.

Materials and Methods

2.1. Animals

Male and female 3xTg-AD mice carry homozygous mutations for presenilin (Psen1), human amyloid (APPSwe), and human tau peptide (tauP301L) transgenes (47) (B6;129-Tg(APPSwe,tauP301L) 1Lfa Psen1^tm1Mpm/Mmjax; MMRRC stock #34830; The Jackson Laboratories, Bar Harbor, ME). Mice were housed in a temperature-controlled room with a 12-hour light/dark cycle and ad libitum access to water and a semi-purified fixed-nutrient diet (AIN-93G, TD.94045, Teklad Envigo, Indianapolis, IN) (71). Animals were mated overnight and the morning of presence of a vaginal plug was designated as embryonic day (E) 0.5. Dams were randomly assigned to an exposure group on E8.5.

Alcohol-exposed (ALC) dams received one daily oral gavage of 3g/kg alcohol (8 dams) (ALC, USP grade 200 proof alcohol, Decon Labs, King of Prussia, PA). Controls (CON) received isocaloric maltodextrin (10 dams) (Envigo Teklad) from E8.5-E17.5 (19,63,72–74). Dams were weighed daily from E0.5-E17.5 with birth typically on E19.5. Offspring were weaned on postnatal day 21 and group-housed with 2–4 same-sex littermates. Mice were weighed monthly. Both sexes (1–2 per litter) underwent behavioral testing at 3mo, 7mo, and 11mo: n=8 for ALC males at 11mo, n=10 for all other groups. All animal protocols were approved by the Institutional Animal Care and Use Committee at the David H. Murdock Research Institute.

2.2. Blood alcohol concentration analysis

To estimate fetal alcohol exposure, blood alcohol concentrations (BACs) were determined in adult females. Mice received a single oral gavage of 3g/kg alcohol. Saphenous vein blood was drawn between 15 and 240 minutes after gavage and quantified using a GM7 Micro-Stat Analyzer (Analox Instruments, Lunenburg, MA).

2.3. Exploration and novel object recognition (NOR)

Male and female offspring were placed in a rectangular arena and allowed to acclimate for 5 minutes (19). Total distance moved during the acclimation period was quantified (Ethovision XT, Noldus Information Technology, Leesburg, VA). After acclimation, two identical objects were introduced into the arena and mice were allowed to explore for 5–10 minutes until they explored the objects for a minimum of 10 seconds (10s). Twenty-four hours later, mice were placed in the same arena with one familiar object and one novel object and left to explore for 3–9 minutes until achieving 10s of object exploration. Exploration time was defined as time the nose-point was within 2 cm of the object perimeter. Percent time exploring the novel object was calculated as (time exploring novel object / total time exploring objects) × 100.

2.4. Auditory fear conditioning

Auditory cued fear conditioning (FC) was performed as previously (19,72,75). On day 1 (acquisition), the mouse was placed in a conditioning chamber (Video Fear Sound Attenuating Cubicles, Med Associates, Fairfax, VT) for 2 minutes, then a sound cue (conditioned stimulus (CS), 2.8 kHz/85 decibel) was played for 30s with a concurrent mild (0.75mA) foot shock (unconditioned stimulus, US) for the last 2s. This CSUS sequence occurred 4 times with an inter-trial interval of 80s. On day 2 (contextual fear), the mouse was placed in the same chamber for 5 minutes with no sound or shock. On day 3 (cued fear), the mouse was placed in a chamber with a different floor, wall pattern, and vanilla odor for 5 minutes. After 2 minutes the CS played for 3 minutes. Percent time freezing (i.e., lack of movement minus respiration) was recorded using VFreeze Software (Med Associates).

2.5. Body composition analysis

Whole body composition was quantified at 12mo (after completion of 11mo behavior testing) using an MRI scanner (EchoMRI^™-100H Body Composition Analyzer, Houston, TX) that quantified total fat mass and lean mass. Percent body fat and percent lean mass were calculated using body weight measured on the same day.

2.6. Glucose tolerance test

Following body composition analysis, the intraperitoneal glucose tolerance test was done as previously (19,63). Mice were fasted for 5 hours then given an intraperitoneal injection of 1.0 g/kg body weight glucose (10% glucose in 0.1M Phosphate Buffered Saline (PBS)). Blood glucose was measured prior to injection (fasting glucose) and at 15, 30, 60, 90, and 120 minutes using a OneTouch Ultra 2 glucometer (LifeScan Inc., Milpitas, CA). Glucose clearance was analyzed by calculating the area under the curve (AUC) using the trapezoidal method and normalizing the baseline to zero (19,76,77).

2.7. Tissue collection

Mice were euthanized by cardiac puncture under isoflurane anesthesia. Brains were removed and weighed. The right hemisphere was fixed overnight in 4% paraformaldehyde in 0.1M PBS, cryoprotected in sucrose, and sectioned at 20μm in the coronal plane.

2.8. Immunohistochemical staining

Sections containing hippocampus were labeled with an antibody directed against the ionized calcium-binding adaptor protein-1 (Iba1) to identify microglia using a procedure similar to that described previously (78). Sections were post-fixed in 4% paraformaldehyde and underwent antigen retrieval in 10mM sodium citrate buffer (pH 6.0). Endogenous peroxidase was inactivated using 3% hydrogen peroxide. Sections were blocked with 2% normal goat serum and 0.1% Triton-X in PBS, then incubated with the primary antibody (1:500 for 2 hours; 019–19741, Wako Chemicals, Richmond, VA) and biotin-conjugated secondary antibody (1:400 for 1 hour, B8895, Sigma Aldrich, St. Lous, MO). Antibody was visualized using peroxidase-conjugated avidin (1:200, PK-6100, Vector Laboratories, Burlingame, CA) followed by 3,3’-diaminobenzidine (DAB, D3939, Sigma-Aldrich, St. Louis, MO). Controls omitted the primary antibody; no labeling was seen.

2.9. Iba1 quantification

Iba1-positive microglia in the hippocampal CA1 and dentate gyrus (DG) were analyzed using a BIOQUANT Image Analysis System (BIOQUANT Image Analysis Corporation, Nashville, TN; Supplemental Figure S1). Four or five sections were quantified from each brain. Labeled cells were identified using the BIOQUANT threshold feature. Total cell size was the area within a perimeter drawn around the tips of the processes for that cell (79). Cell body area was quantified. The ratio of cell body to total cell size was calculated, where higher ratios are indicative of microglial activation.

2.10. Statistical analysis

A one-way analysis of variance (ANOVA) was used to determine differences in dam and litter outcomes. Body composition, fasting glucose, and AUC were analyzed using a two-way ANOVA (sex, exposure). For body weight, NOR, and FC, data were analyzed using a two-way (sex, exposure) repeated measures ANOVA. FC acquisition data were analyzed using a four-way ANOVA (exposure, sex, age, CSUS phase). Iba1 quantification was analyzed using a three-way ANOVA (sex, exposure, region). Two-way ANOVAs were used to analyze differences within a region. Significant differences were followed by post-hoc comparisons with Bonferroni corrections. Statistical significance was set at alpha ≤0.05. Trends were determined when 0.051≤ p≤0.10. Data are presented as either mean ± SEM or mean ± SD. Grubbs’ test (GraphPad Software, San Diego, CA) was used to determine outliers; data from one ALC male was removed from FC. To generate novel hypotheses, exploratory Pearson Correlation analyses were performed to assess associations among the metabolic, behavioral, and microglial outcomes. Because of the large number of comparisons, false discovery rate (FDR) correction was applied (80). Statistics were performed using SPSS (IBM SPSS Statistics, version 27, Armonk, NY) or SigmaPlot (Systat Software Inc., version 14.0, Palo Alto, CA).

3. Results

3.1. Pregnancy outcomes and BACs

Alcohol did not affect gestational weight gain from E0.5-E17.5 or during the gavage window (E8.5-E17.5; F_1,13=0.278; p=0.607) (Table 1). Litter size was also unaffected (F_1,15=0.679; p=0.423), however, percent male was lower in ALC litters compared to CON (F_1,15=7.517; p=0.015). At 30 min BACs were 219 ± 85 mg/dL (Figure 1), similar to that seen in C57BL6/J (B6/J) mice using the same protocol (63,72).

Table 1.

Gestational weight gain, weight gain during the gavage period, litter size, and sex distribution of 3xTg-AD mice. n=8 CON and 10 ALC

	Weight gain (g) E0.5-E17.5	Weight gain (g) E8.5-E17.5	Litter size	% Male
CON	11.84 ± 0.71	9.01 ± 0.81	5.9 ± 1.9	69.8 ± 16.0
ALC	11.98 ± 0.82	9.60 ± 0.62	5.3 ± 0.8	46.4 ± 19.1*

^*Significantly different to CON (p< 0.05). ALC – alcohol-exposed. CON – control. E – embryonic day.

Figure 1. Blood alcohol concentration.

Adult female 3xTg-AD mice received a single oral gavage of 3.0 g/kg alcohol. Blood samples taken from the saphenous vein between 15 and 240 minutes after gavage were analyzed for alcohol content using a GM7 Micro-Stat Analyzer. Values are mean ± SD, n=6–10.

3.2. Offspring body weight and composition

Offspring body weight showed significant interactions between sex and exposure (F_1,755=43.564; p<0.0001, Figure 2) and sex and age (F_19,755=17.003; p<0.0001). Main effects were also seen for sex (F_1,755=52.422; p<0.0001), exposure (F_19,755=74.964; p<0.0001), and age (F_1,755=234.450; p<0.0001). Male body weights were significantly higher than female weights between 1mo and 5mo. Body weights of ALC females continued to increase with age and were higher than CON females beginning at 5mo (F_1,360=80.588; p<0.0001), and surpassed those of ALC and CON males at 9mo.

Figure 2. Body weights and body composition.

(A) Body weights of all offspring increased with age (p<0.0001) and showed sex differences (p<0.0001); male body weights were higher than females between 1mo and 5mo. ALC females had higher body weights than CON females beginning at 5 months (arrow). Prenatal alcohol exposure did not affect male body weights. Values are mean ± SD. (B) Mice underwent MRI body composition analysis at 12mo. Prenatal alcohol exposure did not affect body composition. Females had a higher percent body fat than males and males had a higher percent lean mass than females. Bars are mean ± SD, dots are individual datapoints. n=8–10 per group. Exposure differences indicated by *, sex differences indicated by # (p<0.05). ALC – alcohol-exposed; CON – control; F – female; M – male; mo – months of age.

Two-way ANOVA identified that females had a significantly higher percent body fat at 12mo than males (F_1,34=161.084; p<0.0001; Figure 2). There was a trend for a sex by exposure interaction (F_1,34=3.178; p=0.084); percent fat was higher in ALC females than CON females but lower in ALC males than CON males.

Brain weight was not significantly different among the groups; however the brain weight-to-body weight ratio showed a main effect of exposure (F_1,36=6.718; p=0.014) and sex (F_1,36=7.717; p=0.009; Table 2). The ratio was lower in CON mice than ALC, and higher in males than females.

Table 2.

Brain and body weight of 3xTg-AD mice at 11 months. n = 10

	Brain Weight (g)	Body Weight (g)	Brain weight : Body weight
CON Male	0.47 ± 0.01	35.77 ± 1.24	0.010 ± 0.001
ALC Male	0.46 ± 0.01	37.36 ± 1.01	0.012 ± 0.001
CON Female	0.46 ± 0.01	37.39 ± 1.88	0.013 ± 0.001
ALC Female	0.44 ± 0.01	43.53 ± 1.45*	0.010 ± 0.004*

^*Significantly different to CON (p<0.05). ALC – alcohol-exposed. CON – control.

3.3. Glucose tolerance

A trend for a sex by exposure interaction was observed for fasting blood glucose concentrations (F_1,36=3.820 p=0.058; Figure 3); levels were lower in ALC males than CON males and higher in ALC females than CON females. In the intraperitoneal glucose tolerance test, there was a sex by time interaction (F_1,216=3.115; p=0.010; Figure 3) and a main effect of sex (F_1,216=4.219; p=0.043); blood glucose was higher in females than males at 15 min (p=0.012) and 30 minutes (p=0.003). By 60 minutes, no detectable differences were observed among the groups. There were no significant differences among the groups for AUC (Figure 3).

Figure 3. Intraperitoneal Glucose Tolerance Test (IPGTT).

(A) Fasting blood glucose concentrations showed a trend for sex by exposure interaction; this was lower in ALC males than CON and higher in ALC females than CON females. (B) Blood glucose concentrations in the IPGTT showed a sex by time interaction. Females had higher glucose levels than males at 15- and 30-minutes post-injection. (C) Glucose clearance, determined by area under the curve (AUC) in the IPGTT, was not different among the groups. Values are mean ± SD, dots in A and C are individual datapoints. n=10 per group. Sex differences indicated by # (p<0.05). ALC – alcohol-exposed; AUC – area under the curve for blood glucose levels; CON – control; F – female; M - male.

Using data from all animals, Pearson correlation identified significant positive associations between body weight and percent body fat and between fasting glucose and AUC (Figure 4, Table 3). It also identified positive associations between body weight and AUC and percent body fat and AUC. Negative associations were seen between percent lean mass and body weight and between percent lean mass and AUC.

Figure 4. Pearson correlation analysis among Metabolic Outcomes.

Correlations are positive for all mice between (A) body weight and percent body fat, (B) fasting glucose and AUC, (C) body weight and AUC, and (D) percent body fat and AUC. Symbols represent individual animals; grey symbols are CON, white symbols are ALC. n=8–10 per group. ALC – alcohol-exposed; AUC – area under the curve for blood glucose levels; CON – control; F – female; M – male.

Table 3.

Pearson correlation analysis among metabolic and behavioral outcomes for all animals. P values were adjusted for multiple comparisons using False Discovery Rate (FDR). Yellow highlight indicates significant (p<0.05). Blue highlighting indicates trends (0.05< p<0.10). ALC – alcohol-exposed. AUC – area under the curve for glucose tolerance test. CA1 – cornu Ammonis 1 region of the hippocampus. CON – control. Context – Percent time freezing in contextual fear. Cue – Percent time freezing in cued fear. Distance – total distance moved during exploration. DG – dentate gyrus. Glucose – fasting glucose. NOR – % time exploring novel object in novel object recognition test. P_adj – adjusted p value using FDR.

ALL Animals		AUC	Weight	% Fat	% Lean	NOR	Distance	Context	Cue	DG Iba1	CA1 Iba1
Significant (p<0.05)	Glucose	0.628	0.156	0.0151	−0.00189	−0.0345	−0.0214	0.14	−0.0424	−0.231	−0.125	Correlation Coefficient
Trend (0.5<p<0.1)	1	1.43E-05	0.349	0.928	0.991	0.832	0.896	0.39	0.795	0.182	0.473	P
		0.000197	0.710926	1.0208	0.991	0.973617	1.005714	0.739655	0.99375	0.476667	0.812969	Padj
		40	38	38	38	40	40	40	40	35	35	N
	AUC		0.368	0.514	−0.497	−0.23	−0.0868	0.28	0.157	−0.0558	0.00951	Correlation Coefficient
			0.023	0.000957	0.00151	0.154	0.594	0.0797	0.334	0.75	0.957	P
			0.097308	0.005848	0.008305	0.445789	0.859737	0.273969	0.706538	0.982143	1.012212	Padj
			38	38	38	40	40	40	40	35	35	N
	Weight			0.65	−0.632	−0.256	−0.156	0.295	0.538	0.00289	0.118	Correlation Coefficient
				1.01E-05	2.05E-05	0.121	0.349	0.0719	0.000494	0.987	0.508	P
				0.000185	0.000226	0.369722	0.685536	0.282464	0.003881	1.005278	0.821765	Padj
				38	38	38	38	38	38	34	34	N
	% Body Fat				−0.987	−0.204	0.139	0.438	0.566	−0.0484	−0.0142	Correlation Coefficient
					2.50E-30	0.22	0.405	0.00593	0.000214	0.786	0.936	P
					1.37E-28	0.55	0.7425	0.02965	0.001962	1.005349	1.009412	Padj
					38	38	38	38	38	34	34	N
	% Lean					0.17	−0.177	−0.408	−0.536	0.045	0.0262	Correlation Coefficient
						0.308	0.289	0.011	0.00053	0.801	0.883	P
						0.705833	0.691087	0.050417	0.003644	0.979	1.011771	Padj
						38	38	38	38	34	34	N
	NOR						0.065	0.0938	−0.218	−0.282	−0.302	Correlation Coefficient
							0.69	0.565	0.176	0.101	0.0782	P
							0.94875	0.863194	0.484	0.326765	0.286733	Padj
							40	40	40	35	35	N
	Distance							−0.0579	0.121	0.0425	0.00948	Correlation Coefficient
								0.723	0.458	0.809	0.957	P
								0.969878	0.812581	0.967283	0.993113	Padj
								40	40	35	35	N
	Context								0.164	−0.0968	−0.12	Correlation Coefficient
									0.312	0.58	0.494	P
									0.6864	0.862162	0.823333	Padj
									40	35	35	N
	Cue									0.0712	0.11	Correlation Coefficient
										0.684	0.53	P
										0.964615	0.832857	Padj
										35	35	N
	DG Iba1										0.94	Correlation Coefficient
											5.88E-17	P
											1.62E-15	Padj
											35	N

3.3. Behavior

Novel Object Recognition:

PAE did not affect the exploratory behavior of mice during habituation to the arena, however distance traveled was greater at 3mo than at 7mo or 11mo in both sexes and exposure groups (F_2,106=6.931, p=0.001; Figure 5). There was a significant group x age interaction on percent time exploring the novel object (F_2,106=5.274, p=0.007; Figure 5). CON animals outperformed ALC at 11mo, and CON animals spent more time exploring the novel object at 11mo than at 3mo.

Figure 5. Exploration and Novel Object Recognition.

(A) Mice were tested on exploratory behavior in an empty arena at 3, 7, and 11mo. Total distance traveled decreased with age in all mice. (B) Prenatal alcohol exposure altered recognition memory in an age-dependent manner. CON mice spent more time exploring the novel object than ALC mice at 11mo, and CON animals spent more time exploring the novel object at 11mo than 3mo. Values are mean ± SEM, n=8–10 per group, dots are individual data points. Age differences indicated by §, exposure differences indicated by * (p <0.05). ALC – alcohol-exposed; CON – control; F – female; M – male; mo – months of age.

Fear Conditioning:

Acquisition. There were interactions between exposure, sex, and age on percent time freezing to each CSUS presentation (CSUS#; F_2,515=4.728, p=0.009; Figure 6) and sex, age, and CSUS# (F_8,515=2.131, p=0.031). CON and ALC females and ALC males froze less at 3mo than 7mo or 11mo. Females froze more than males during exploration, CSUS2, and CSUS4 at 7mo and during exploration and CSUS1 at 11mo. Significant main effects were seen for group (ALC froze more than CON; F_1,515=5.755, p=0.017), sex (females froze more than males; F_1,515=27.007, p<0.001), age (freezing was lower at 3mo than 7mo or 11mo; F_1,515=20.878, p<0.001), and phase (freezing increased significantly across successive phases except between CSUS3 and CSUS4; F _1,515=78.122, p<0.001).

Figure 6. Auditory cued fear.

During acquisition, mice froze less at 3mo (A) than 7 (B) or 11mo (C). (B) Females froze more than males during initial exploration (0 cue/shock presentation) and cue/shock presentations 2 and 4 at 7mo. (D) Females froze more during exploration and CSUS1 at 11mo. (D) Females froze more than males in the context, and all animals froze less at 11mo than 3mo or 7mo. (E) ALC mice froze more than CON in response to the cue and females froze more than males. Sex differences indicated by #, age differences indicated by §, exposure differences indicated by * (p < 0.05). Values are mean ± SEM, dots show individual datapoints. n=7–10 per sex per group. ALC – alcohol-exposed; CON – control; CSUS – conditioned stimulus paired with unconditioned stimulus; F – female; M – male; mo – months of age.

Context.

Analysis of mean percent time freezing in the context only identified main effects of sex (F_1,103=4.337, p=0.04; Figure 6) and age (F_2,103=11.824, p<0.0001). Females froze more than males and animals froze less at 11mo than at 3mo or 7mo. Cue. Mean percent time freezing to the cue showed effects of exposure (F_1,103=5.902, p=0.017; Figure 6) and sex (F_1,103=10.643, p=0.001). ALC mice froze more than CON, and females froze more than males.

Using data from all animals, Pearson correlations identified associations among the metabolic and behavior outcomes (Figure 7, Table 3): these were most apparent in FC. Percent time freezing in context positively associated with percent body fat and negatively with percent lean mass. Percent time freezing to the cue positively associated with body weight and percent body fat, and negatively with percent lean mass. Analysis of subgroups identified significant positive associations between percent body fat and freezing in context or to the cue in ALC; these were negative for percent lean mass (Figure 7; Table 4). In CON animals, there were trends for associations between percent body fat and freezing to the cue (Figure 7; Table 5). Associations between freezing in context and body weight also showed a positive correlation in ALC animals and freezing to cue and body weight was positively correlated in both ALC and CON, although none of these outcomes survived FDR correction.

Figure 7. Pearson correlation analysis between metabolic and fear conditioning outcomes.

Using data from all animals there are positive correlations between (A) % body fat and % time freezing in the context, (B) body weight and % time freezing to the cue, and (C) % body fat and % time freezing during cue learning. Analysis of subgroups identified that the associations were apparent in ALC but not CON animals (Tables 4 and 5). Symbols represent individual animals; grey shapes and solid line are CON, white shapes and dashed line are ALC. n=8–10. ALC – alcohol-exposed; CON – control; F – female.

Table 4.

Pearson correlation analysis among metabolic and behavioral outcomes for ALC animals. P values were adjusted for multiple comparisons using False Discovery Rate (FDR). Yellow highlight indicates significant (p<0.05). Blue highlighting indicates trends (0.05< p<0.10). ALC – alcohol-exposed. AUC – area under the curve for glucose tolerance test. CA1 – cornu Ammonis 1 region of the hippocampus. CON – control. Context – Percent time freezing in contextual fear. Cue – Percent time freezing in cued fear. Distance – total distance moved during exploration. DG – dentate gyrus. Glucose – fasting glucose. NOR – % time exploring novel object in novel object recognition test. P_adj – adjusted p value using FDR.

ALC Animals		AUC	Weight	% Fat	% Lean	NOR	Distance	Context	Cue	DG Iba1	CA1 Iba1
Significant (p<0.05)	Glucose	0.684	0.553	0.365	−0.339	−0.00415	−0.243	0.463	0.169	−0.473	−0.43	Correlation Coefficient
Trend (0.5<p<0.1)	1	0.000879	0.0173	0.136	0.169	0.986	0.303	0.0399	0.477	0.0551	0.0851	p
		0.006043	0.067964	0.325217	0.3575	0.986	0.574655	0.129088	0.84629	0.1595	0.21275	Padj
		20	18	18	18	20	20	20	20	17	17	N
	AUC		0.77	0.669	−0.622	−0.272	−0.323	0.428	0.479	−0.116	−0.054	Correlation Coefficient
			0.000188	0.00238	0.00589	0.246	0.165	0.0598	0.0324	0.659	0.837	P
			0.002068	0.014544	0.032395	0.483214	0.363	0.16445	0.111375	1.035571	0.979468	Padj
			18	18	18	20	20	20	20	17	17	N
	Weight			0.805	−0.798	−0.0767	−0.309	0.523	0.577	−0.163	−0.083	Correlation Coefficient
				5.67E-05	7.21E-05	0.762	0.212	0.0259	0.0122	0.548	0.76	P
				0.00104	0.000991	1.074615	0.431852	0.094967	0.051615	0.913333	1.1	Padj
				18	18	18	18	18	18	16	16	N
	% Fat				−0.994	0.0313	0.0583	0.593	0.767	0.0168	0.0374	Correlation Coefficient
					1.39E-16	0.902	0.818	0.00953	0.000204	0.951	0.891	P
					7.66E-15	0.972745	0.999778	0.04765	0.00187	0.986887	0.9801	Padj
					18	18	18	18	18	16	16	N
	% Lean					−0.0393	−0.0707	−0.578	−0.736	−0.0497	−0.0619	Correlation Coefficient
						0.877	0.78	0.0121	0.000498	0.855	0.82	P
						0.984388	0.997674	0.055458	0.003913	0.979688	0.980435	Padj
						18	18	18	18	16	16	N
	NOR						0.455	0.406	−0.0709	−0.186	−0.178	Correlation Coefficient
							0.044	0.0759	0.766	0.475	0.495	P
							0.134444	0.198786	1.05325	0.870833	0.850781	Padj
							20	20	20	17	17	N
	Distance							0.103	0.0064	0.0774	−0.064	Correlation Coefficient
								0.666	0.979	0.768	0.807	P
								1.0175	0.99713	1.030244	1.00875	Padj
								20	20	17	17	N
	Context								0.325	−0.125	−0.104	Correlation Coefficient
									0.162	0.632	0.692	P
									0.37125	1.022353	1.028649	Padj
									20	17	17	N
	Cue									0.0197	0.0763	Correlation Coefficient
										0.94	0.771	P
										0.994231	1.009643	Padj
										17	17	N
	DG Ibal										0.968	Correlation Coefficient
											1.96E-10	P
											5.39E-09	Padj
											17	N

Table 5.

Pearson correlation analysis among metabolic and behavioral outcomes for CON animals. P values were adjusted for multiple comparisons using False Discovery Rate (FDR). Yellow highlight indicates significant (p<0.05). Blue highlighting indicates trends (0.05< p<0.10). ALC – alcohol-exposed. AUC – area under the curve for glucose tolerance test. CA1 – cornu Ammonis 1 region of the hippocampus. CON – control. Context – Percent time freezing in contextual fear. Cue – Percent time freezing in cued fear. Distance – total distance moved during exploration. DG – dentate gyrus. Glucose – fasting glucose. NOR – % time exploring novel object in novel object recognition test. P_adj – adjusted p value using FDR.

CON Animals |		AUC	Weight	% Fat	% Lean	NOR	Distance	Context	Cue	DG Iba1	CA1 Iba1
Significant (p<0.05)	Glucose	0.583	−0.0574	−0.295	0.297	−0.199	0.133	−0.074	−0.173	0.215	0.385	Correlation Coefficient
Trend (0.5<p<0.1)		0.00702	0.81	0.206	0.203	0.4	0.576	0.756	0.465	0.392	0.114	P
		0.1287	0.891	0.539524	0.55825	0.733333	0.772683	0.848571	0.730714	0.743448	0.482308	Padj
		20	20	20	20	20	20	20	20	18	18	N
	AUC		0.0434	0.346	−0.366	−0.291	0.17	0.171	−0.195	0.0256	0.12	Correlation Coefficient
			0.856	0.135	0.113	0.213	0.473	0.47	0.411	0.92	0.636	P
			0.923137	0.436765	0.517917	0.5325	0.703108	0.718056	0.729194	0.937037	0.813488	Padj
			20	20	20	20	20	20	20	18	18	N
	Weight			0.512	−0.498	−0.19	−0.076	0.185	0.468	−0.251	0.028	Correlation Coefficient
				0.0211	0.0254	0.424	0.75	0.435	0.0374	0.315	0.912	P
				0.290125	0.2794	0.706667	0.87766	0.703676	0.342833	0.693	0.946415	Padj
				20	20	20	20	20	20	18	18	N
	% Fat				−0.981	−0.44	0.207	0.35	0.41	−0.375	−0.229	Correlation Coefficient
					3.20E-14	0.0522	0.382	0.131	0.0728	0.125	0.361	P
					1.76E-12	0.410143	0.778148	0.450313	0.5005	0.458333	0.763654	Padj
					20	20	20	20	20	18	18	N
	% Lean					0.389	−0.261	−0.306	−0.388	0.385	0.269	Correlation Coefficient
						0.0897	0.267	0.189	0.0913	0.115	0.281	P
						0.548167	0.638478	0.547105	0.50215	0.451786	0.643958	Padj
						20	20	20	20	18	18	N
	NOR						−0.315	−0.101	−0.191	0.399	0.167	Correlation Coefficient
							0.177	0.672	0.42	0.101	0.507	P
							0.540833	0.803478	0.721875	0.505	0.733816	Padj
							20	20	20	18	18	N
	Distance							−0.138	0.206	0.115	0.147	Correlation Coefficient
								0.563	0.384	0.651	0.56	P
								0.774125	0.754286	0.81375	0.789744	Padj
								20	20	18	18	N
	Context								0.0746	−0.0192	−0.111	Correlation Coefficient
									0.754	0.94	0.661	P
									0.863958	0.94	0.807889	Padj
									20	18	18	N
	Cue									−0.12	−0.0461	Correlation Coefficient
										0.635	0.856	P
										0.831548	0.905385	Padj
										18	18	N
	DG Ibal										0.794	Correlation Coefficient
											8.22E-05	P
											0.002261	Padj
											18	N

3.4. Iba1 Morphology

The ratio of cell body to total cell size in the hippocampus was affected by exposure (F_1,62= 23.240; p<0.0001) and showed a trend for region (F_1,62=3.099; p=0.083). For both regions, the ratio was larger in ALC than CON, suggesting that PAE caused changes indicative of microglial activation, specifically, larger cell bodies and/or shorter processes (Figure 8).

Figure 8. Microglial activation in the hippocampus.

Iba1 staining of microglia in the CA1 and DG regions of hippocampus in 12mo CON and ALC brains at low power (5x; A and B) and higher power (20x; C-F). Non-activated Iba1+ microglia cells (arrow) have a smaller cell body and longer thinner processes encompassing a larger area in CON CA1 (C) and DG (E) regions. In ALC brains, Iba+ microglia have a more amoeboid morphology (arrowhead) suggestive of microglial activation in CA1 (D) and DG (F) regions. The ratio of cell body to total cell area region was higher in ALC mice than CON in the CA1 (G) and DG (H). Exposure differences indicated by * (p<0.05). Values are mean ± SD, dots show individual datapoints. n=8–9 per sex per group. ALC – alcohol-exposed; CON – control; DG – dentate gyrus; F – female; M – male; mo – months of age.

Using data from all animals, Pearson correlation identified a significant correlation between the ratio in CA1 with that in DG (Figure 9, Table 3) and a trend for negative correlations between the cell body to total cell size ratio for CA1 with % time exploring the novel object. Associations of the ratios with metabolic or fear conditioning outcomes were not significant.

Figure 9. Association analysis between microglial activation and behavioral outcomes.

Linear regression using data from all animals shows trends for negative correlations between percent time exploring the novel object and a measure of microglial activation in the CA1 region (A). (B) A significant positive correlation between microglial activation in the CA1 region and the DG region was observed. Symbols represent individual animals; grey shapes are CON, white shapes are ALC. n=8–10. ALC – alcohol-exposed; CON – control; DG – dentate gyrus; F – female; M - male.

4. Discussion

We tested the hypothesis that PAE exacerbates the cognitive decline associated with AD and that this was associated with greater metabolic impairment and microglial activation in a mouse model of AD. To the best of our knowledge, this study is the first to investigate effects of PAE in a model of AD. We found that PAE caused age-dependent behavioral deficits compared to controls and these deficits were accompanied by metabolic dysfunctions with respect to glucose tolerance and adiposity in ALC mice. Importantly, the metabolic dysfunctions strongly correlated with observed behavioral deficits. These correlations were exposure-dependent and strongest in females. In AD, metabolic dysfunction correlates with cognitive decline (81). Our data suggests that PAE may be contributory to these associations, and we note that between 5% and 13% of the U.S. population experience PAE (1). Brains from ALC mice were characterized by a higher cell body to cell size ratio in hippocampal microglia; such morphology is associated with microglial activation and neuroinflammation. Intriguingly, we identified a trend for microglial activation to negatively correlate with recognition memory. These changes were evidenced using an alcohol exposure that models the Alcohol-Related Neurodevelopmental Disorders, a ‘milder’ outcome that includes cognitive and behavioral deficits but no birth weight reduction or facial deficits. Collectively, our findings suggest that PAE could progress age-related diseases and supports the notion that FASD is a whole-body disorder.

Cognitive behaviors were tested in 3xTg-AD mice at three different ages in adulthood including an age prior to pathological changes in this model (3mo), an age around the emergence of the pathological changes (7mo), and a later age where cognitive impairment is more advanced (11mo) (48,70). Both CON and ALC animals showed some evidence of an age-related cognitive impairment in the cued fear conditioning task, with less freezing to the context at 11mo than at younger ages: these findings are consistent with prior 3xTg-AD studies (82–85). PAE affected the response to the cue in the fear conditioning, such that ALC females were hyper-responsive at 7mo and 11mo. This was also seen in adolescent, but not aged, B6/J mice exposed to alcohol prenatally (19) and in 7mo 3xTg-AD mice exposed to alcohol during adulthood (85). A PAE-induced deficit in recognition memory was seen at 11mo. Specifically, CON animals showed greater recognition memory at 11mo than at younger ages, but ALC did not. A similar training-related improvement in outcome was previously reported for 3xTg-AD mice in the Morris water maze, suggesting they can learn and remember (86). It is possible that repeated testing is a form of enrichment that is protective against later behavioral and/or brain changes (48,86). At this time, we cannot determine if the improved performance in recognition memory is a result of the re-test model or an age-effect. Regardless, ALC mice do not benefit from this.

Other animal studies have reported that PAE worsens glucose handling and increases adiposity in age- and dose-related manners (14,19), and clinical studies suggest this is also observed in FASD (6,7,10). We previously reported that, using this same exposure, ALC B6/J females had higher body weights than CON animals beginning around 7.5mo and worsened glucose handling at 17mo (19). A similar outcome occurs in this 3xTg-AD model, and interestingly, we find these traits emerge earlier here than in the B6/J strain. It is possible that the combination of PAE and the genetics of the 3xTg-AD mice plays a role in accelerating their metabolic dysfunction, as the links between obesity, T2D and the risk of AD have been established (87–89).

In the present study, we find correlations between cognitive performance and metabolic outcomes, wherein worse metabolic outcomes correlate with worse behavior. These effects are driven by the ALC animals, likely because their metabolic outcomes tended to be worse than CON animals. Glucose tolerance was only measured at 12mo, therefore, we do not know when it diverged, but it is possible that ALC males may have developed impairments at an older age or if challenged with a high-fat diet (53). In our previous study that tested 20mo old B6/J mice (19), both ALC males and females had greater impairments in glucose tolerance compared with their age-matched CON. The underlying mechanisms for these exposure- and sex-dependent effects on glucose tolerance are not completely understood, and both hepatic and pancreatic dysfunction could be contributory. Both can partially characterize PAE and have been observed in 3xTg-AD females with worse glucose handling and higher amyloid beta levels (10,15,17,18,90–92).

It is possible that the metabolic dysfunction observed in PAE may contribute to the behavioral deficits, specifically, the inability to learn or improve behavior in later adulthood. Age-related diseases, like obesity and T2D, are associated with higher risk of cognitive decline (58,93–97). The underlying mechanisms are incompletely understood at this time, and both chronic inflammation and vascular impairment have been posited as contributory (45,57,98); we note that both can characterize PAE (32,99). Our data suggest an opportunity to evaluate if treatment for metabolic dysfunctions can either improve behavioral outcomes or prevent their deterioration, and it is worth considering whether the difficulties experienced by adults with FASD in accessing health care may have larger implications for their cognitive and behavioral challenges (6).

One of the most important findings of this study is that the changes in microglial morphology that partially characterize PAE and contribute to neuroinflammation (28–31) were present at 12mo in the hippocampus of both sexes, suggesting this effect is long-lasting. The 3xTg-AD mice have been genetically altered to express humanized presenilin, tau, and amyloid beta. Amyloid beta and tau pathology is seen in the CA1 region of hippocampus as early as 6mo (47) and is present throughout hippocampus in all 3xTg-AD females by 12mo, and inconsistently in males (52). It is possible that the microglia respond to these pathological changes, as microglial activation was also seen in 6mo 3xTg-AD female hippocampus (52) and Iba-1 protein is elevated in the hippocampus of 3xTg mice compared with non-transgenic animals at 8mo (100). In the present study, ALC animals showed significantly higher levels of activation ~12mo after the last exposure to alcohol despite CON and ALC mice being housed in neighboring cages, and with the same food, water, and bedding. Others report persistent neuroinflammation in models of PAE as old as 3mo (28,101). Our data suggest that microglia may remain activated through adulthood and could further contribute to the pathophysiology of FASD. Other cell types, including astrocytes and endothelial cells, also play a role in neuroinflammation (102) and may contribute to the phenotype. It is notable that the morphology showed trends to negatively correlate with recognition memory regardless of sex or exposure. These outcomes did not survive FDR correction, likely due to the variability inherent in behavior data, however, this is an intriguing finding that could be evaluated in an independent study.

In summary, we show that adult mice that experienced PAE show cognitive impairments as they age, and that PAE does not appear to interact with AD risk to further worsen these deficits with age. These cognitive deficits strongly correlate with the animal’s worsening adiposity and glucose intolerance, and this association is strongest in the ALC females. These findings suggest opportunities for intervention that may mitigate these outcomes for FASD.