Butyl benzyl phthalate as a key component of phthalate ester in relation to cognitive impairment in NHANES elderly individuals and experimental mice

Phthalates are a group of neurotoxicants with cognitive-disrupting potentials. Given the structural diversity of phthalates, the corresponding neurotoxicity is dramatically altered. To identify the potential contributions of different phthalates on the process of cognitive impairment, data of 836 elders from the NHANES 2011–2014 cycles were used. Survey-weighted logistic regression and principal component analysis-weighted quantile sum regression (PCA-WQSR) models were applied to estimate the independent and combined associations of 11 urinary phthalate metabolites with cognitive deficit (assessed by 4 tests: Immediate Recall (IR), Delayed Recall (DR), Animal Fluency (AF), and Digit Symbol Substitution Test (DSST)) and to identify the potential phthalate with high weight. Laboratory mice were further used to examine the effect of phthalates on cognitive function and to explore the potential mechanisms. In logistic regression models, MBzP was the only metabolite positively correlated with four tests, with ORs of 2.53 (quartile 3 (Q3)), 2.26 (Q3), 2.89 (Q4) and 2.45 (Q2), 2.82 (Q4) for IR, DR, AF, and DSST respectively. In PCA-WQSR co-exposure models, low-molecular-weight (LMW) phthalates were the only PC positively linked to DSST deficit (OR: 1.93), which was further validated in WQSR analysis (WQS OR7-phthalates: 1.56 and WQS OR8-phthalates: 1.55); consistent with the results of logistic regression, MBzP was the dominant phthalate. In mice, butyl benzyl phthalate (BBP), the parent phthalate of MBzP, dose-dependently reduced cognitive function and disrupted hippocampal neurons. Additionally, the hippocampal transcriptome analysis identified 431 differential expression genes, among which most were involved in inhibiting the neuroactive ligand-receptor interaction pathway and activating the cytokine-cytokine receptor interaction pathway. Our study indicates the critical role of BBP in the association of phthalates and cognitive deficits among elderly individuals, which might be speculated that BBP could disrupt hippocampal neurons, activate neuroinflammation, and inhibit neuroactive receptors. Our findings provide new insight into the cognitive-disrupting potential of BBP.


Introduction
With the sustainable increment of global elderly populations, aging-related disease issues are becoming nonnegligible public health challenges (Partridge et al. 2018). Cognitive impairment is one of the most prominent health conditions in the era of elderly individuals and is characterized by impaired mental abilities to concentrate, learn, reason, remember, and make decisions (Kivipelto et al. 2018;Power et al. 2019). Across different elderly populations, the prevalence of cognitive decline is estimated to range from 5.1 to 41%, with a median of 19.0% (Pais et al. 2020). This kind of structural and functional alteration in the brain may Responsible Editor: Lotfi Aleya result in reduced quality of daily life and an elevated risk of dementia, a hallmark feature of severe cognitive decline, and death (Arai et al. 2018). It was estimated that the counts of individuals suffering from dementia would rise to more than 131 million globally by 2050 (Collaborators 2022). Additionally, cognitive decline was associated with several social problems, such as loss of autonomy and independence and consistent requirements of caregivers and health services (Terada et al. 2019). Hence, the need for investigations on the potential preventive risk factors for cognitive impairment is growing more urgent.
Although cognitive impairment is a general and feared feature of aging, the etiology of cognitive impairment involves complex interactions between genetic factors and modifiable environmental factors . Indeed, by controlling the potential confounders such as age, lifestyle factors, and medical history factors, several epidemical studies have been performed recently to explore the association between environmental pollution and cognitive impairment among older adults (Cao et al. 2021;Przybyla et al. 2017).
Phthalates are a group of multifunctional compounds ubiquitously used as plasticizers and solvents in manufacturing products, such as cosmetics, food wraps, housing materials, and personal care products (Qin et al. 2021). Phthalates are non-persistently bound to polymers, and therefore the constant release or migration of phthalates from consumer products into the environment has raised dramatic public health concerns (Mohammadi et al. 2022). In addition to the expected reproductive disorders (Martínez-Ibarra et al. 2021), phthalate exposure has also been linked to disorders of the nervous system. For instance, prenatal exposure to phthalates is linked to autistic traits in young children, while early childhood exposure is correlated with autistic traits in school-aged children . The levels of urinary phthalate metabolites were positively correlated with the neurotoxic metabolite quinolinic acid in both male and female participants in Massachusetts (Nassan et al. 2019(Nassan et al. , 2020. However, regarding cognitive impairment, studies have mainly focused on the deleterious effects of phthalates on neurodevelopment in children, and human data in adults are rather limited (Radke et al. 2020). To date, only three epidemiological studies have explored the association between phthalates and cognitive function among elderly adults, and their findings are inconsistent (Shiue 2015;Shiue and Starr 2012;Weng et al. 2022).
In real life, individuals are usually exposed to mixtures of several phthalates rather than a single phthalate (Wang et al. 2018). Based on the corresponding molecular weight and length of the carbon chain backbone, environmental phthalates can be classified as high-molecular-weight (HMW) phthalates, such as di-(2-ethylhexyl) phthalate (DEHP), di-iso-nonyl phthalate (DiNP), and di-iso-decyl phthalate (DiDP) and low-molecular-weight (LMW) phthalates, such as butyl benzyl phthalate (BBP), di-iso-butyl phthalate (DiBP), and di-n-butyl phthalate (DBP) . Of note, the neurotoxic patterns and underlying mechanisms of phthalates were remarkably altered due to the structural diversity (Qureshi et al. 2016). As illustrated in an experimental study of zebrafish embryos treated with six phthalates, BBP, DEHP, and DiNP were found to affect larval neurobehaviors and dopaminergic systems, while no obvious alteration was recognized in the dimethyl phthalate (DMP), diethyl phthalate (DEP), or di-n-octyl phthalate (DnOP) exposure group (Tran et al. 2021). Thus, it is essential to identify the potential contributions of different phthalates to the process of cognitive impairment among elderly individuals.
The absorbed phthalates are readily metabolized to their corresponding monoesters or oxidative metabolites and subsequently excreted in the urine (Urbancova et al. 2019). Given the routine and continuous exposure of phthalates in daily life, urinary phthalate metabolites are widely used as appropriate biomarkers of phthalate internal exposure (Lu et al. 2020). Therefore, in the present study, we explored the association of urinary phthalate metabolites with cognitive impairment using data from the National Health and Nutrition Examination Survey (NHANES). Survey-weighted multivariable logistic regression and a two-stage analysis of principal component analysis (PCA) and weighted quantile sum regression (WQSR) were used to assess the contribution of certain phthalate metabolites to the risk of cognitive impairment and identify potential cognitive-toxic candidates. In addition, experimental mice were used to further confirm the deleterious effect of phthalate candidates on cognitive functions. By exploring and validating certain phthalate candidates that involved in the progression of cognitive impairment, our study may provide new insight into phthalate-induced cognitive toxicity in elderly individuals.

Study design and data collection
NHANES is a nationwide population-based health survey with wide-ranging and representative data with respect to environmental exposure, demographic information, and nutritional and health outcomes of noninstitutionalized subjects in the USA. The cross-sectional, multistage, and random sampling design of NHANES makes it a suitable platform for investigators in the risk assessment of environmental pollutants (Sobus et al. 2015). Detailed information on the NHANES database is available on the website: http:// www. cdc. gov/ nchs/ nhanes. htm.
In the current study, we obtained data from two consecutive 2-year cycles (2011-2012 and 2013-2014) where a series of cognitive assessments were performed among adults aged 60 years or older. Overall, a total of 961 elderly subjects with complete data on cognitive tests and urinary phthalate metabolites were initially included. Among them, 125 individuals with missing data regarding covariates, subsample A weight (WTSA2YR), and environmental B 2-year weights (WTSB2YR) were subsequently excluded. The final study enrolled 836 elderly adults and the corresponding flow chart is shown in Supplementary Fig. S1.

Assessments of urinary phthalate metabolites
Urine specimens were collected using a prescreened urine sampling device and stored frozen at − 40 °C or below. The urine specimens were initially processed using enzymatic deconjugation of the glucuronidated analytes. Online solidphase extraction (SPE) coupled with reversed phase high performance liquid chromatography-electrospray ionizationtandem mass spectrometry (HPLC-ESI-MS/MS) was subsequently performed to quantitatively evaluate the concentrations of phthalate metabolites.

Assessments of cognitive function
Cognitive assessments included word learning and recall tests from Consortium to Establish a Registry for Alzheimer's Disease Word Learning subset (CERAD-WL), the Animal Fluency (AF) test, and the Digit Symbol Substitution Test (DSST). The immediate recall module (IR) of CERAD-WL consists of three consecutive learning trials, and for each trial, individuals were first instructed to read 10 unrelated words aloud, one at a time. Immediately after the presentation of the last word on the computer monitor, subjects were then asked to immediately recall as many words as possible. The delayed recall module (DR) of CERAD-WL was conducted after the other two cognitive assessments (AF and DSST) were finished, and the participants were asked to recall the 10 words presented in the first learning trial. The AF test, by asking individuals to name as many animals as possible within 1 min, was performed to evaluate the categorical verbal fluency, an integral component of executive function. The DSST, a paper-based performance module from the Wechsler Adult Intelligence Scale, was used to examine brain function with respect to processing speed, sustained attention, and working memory. The exercise has a set of symbols with a matching key and participants were asked to draw the corresponding symbols in the 133 boxes within 2 min. The total score of each cognitive measurement was 30 for IR, 10 for DR, 40 for AF, and 100 for DSST. In the present study, individuals who scored < 17 for IR, < 5 for DR, < 14 for AF, or < 34 for DSST were considered to have potential cognitive impairment according to prior studies (Bailey et al. 2020).

Covariates
A series of factors that have been previously linked to the cognitive functions or phthalate metabolite levels were included as potential confounders (Iranpour et al. 2020;Yeh et al. 2021). Specifically, we selected and adjusted demographic factors, including age, sex, ethnicity, education level, and family poverty income ratio (PIR); lifestyle factors, including physical activity, smoking and drinking status; medical history factors, including hypertension, diabetes, stroke, and cardiovascular disease (CVD); and urinary creatinine concentration and body measure index (BMI) in the current study. Participants' PIR, BMI, and age were further grouped into two, three, and four levels, respectively: PIR: < 1 and ≥ 1; BMI: < 25 kg/m 2 , 25-30 kg/ m 2 , and ≥ 30 kg/m 2 ; and age: 60-65 years old, 65-70 years old, 70-75 years old, and > 75 years old.
Physical activity was classified as no or lower, moderate, and vigorous based on the questions "Moderate work activity" and "Vigorous work activity." Smoking status was categorized into two groups (nonsmokers and smokers) according to the questions "Smoked at least 100 cigarettes in life" and "Do you now smoke cigarettes." Participants who had smoked at least 100 cigarettes in life and now smoked every day or some days and those who had smoked at least 100 cigarettes in life but now stopped smoking were recognized as smokers, while those who had never smoked at least 100 cigarettes in life were nonsmokers. Drinking status was determined based on the question "Had at least 12 alcohol drinks per year." Hypertension was determined based on the average value of blood pressure tested for four days. The condition of diabetes was calculated according to the question "Doctor told you have diabetes," glycohemoglobin HbA1C (> 6.5%), fasting glucose (≥ 7 mmol/L), random blood glucose (≥ 11.1 mmol/L), 2-h OGTT blood glucose (≥ 11.1 mmol/L), or use of diabetes medication. CVD was determined based on the question "Ever told you had a congestive heart failure, coronary heart disease or angina." Stroke was calculated based on the question "Ever told you had a stroke" in the questionnaire (mcq160f) and examination (spq070d) part of NHANES.

Statistical analysis of epidemiological data
The levels of urinary phthalates were first natural log-transformed due to their skewed distribution. Pearson correlation analysis was applied to explore the correlation coefficients among all the exposure chemicals. Multivariate logistic regression was then adopted to determine the associations between urinary phthalates and cognitive impairment (IR, DR, AF, and DSST) using odds ratios (ORs) and corresponding 95% confidence intervals (CIs). Medians and interquartile ranges were calculated for each phthalate metabolite, and subjects in quartile 2 (Q2), Q3, and Q4 were compared to those in Q1. Tests for trends across categories were performed by fitting phthalate metabolites as continuous variables using the median value of each category. The complex design of the survey and subsample weights were considered, and the 4-year sample weight was computed by dividing the 2-year subsample A or B weight by 2. In the sensitivity analysis, data were reanalyzed without accounting for the subsample weights to test the robustness of the results. To explore the joint effect of multiple phthalates on cognitive impairment and further assess the contribution of different phthalates to dysfunctions, a PCA-WQSR twostage analysis was performed based on a previous study (Tao et al. 2021). Briefly, we first calculated the Kaiser Meyer Olkin (KMO) value of 11 phthalates and conducted Bartlett's test and scree test. A total of 11 principal components (PCs) were extracted, and 3 PCs with eigenvalues > 1 were selected. Varimax rotation was used with Kaiser normalization to better interpret the contribution of each phthalate to the selected PCs. The scores of individuals for selected PCs were divided into 4 quantiles and fitted into our previous logistic regression models to assess the association of chosen PCs and cognitive impairment.
Given the limitation of the PCA model in considering outcomes and identifying vital components, WQSR analysis, a useful method to reduce the high dimensionality and multicollinearity of multiple correlated factors, was subsequently performed to validate the significant results observed in PCA models and evaluate the contribution of each phthalate. In this study, WQSR estimated a weighted linear index of each individual by dividing phthalates into quartiles, which reflected the proportion of contribution to the combined effects. Bootstrap sampling (n = 500) was used to estimate the effect robustly, and the final weighted indices were calculated as the average values across the bootstrap samples. For each WQSR model, the top phthalates in PCs with significant associations were included, and the direction was assumed to be positive or negative according to the results of PCA. An explorative method of the top 6, 7, 8, 9, and 10 phthalates in selected PCs was conducted, as there is no reference or prior experience on the threshold of phthalate selection. Restricted cubic splines (RCSs) with four knots (5th, 35th, 65th, and 95th percentiles) were then carried out based on the significant association of WQS index with cognitive impairment and used to visualize the exposure-response curves of WQS index with cognitive impairment, and the Wald test was applied to evaluate the violations of linearity.
The statistical analyses were conducted using SAS 9.4 (SAS Institute Inc., Cary, NC, USA) with the PROC SUR-VEYREG procedure, which involved the NHANES primary sampling unit, strata, and sampling weights of phthalates. The RCS, PCA, and WQSR analyses were performed with R software (version 4.1.2) using the "rcs," "psych," and "gWQS" packages. A value P ≤ 0.05 was considered statistically significant.

Animals and treatments
Male C57BL/6 mice were obtained from the Animal Core Facility of Nanjing Medical University and maintained in a strictly controlled specific pathogen-free (SPF) environment with a 12-h light/dark cycle at a suitable temperature (25 ± 1 °C) and relative humidity (50-60%). Sterilized food and distilled water were provided ad libitum. All animals were treated humanely, and the experimental procedures in the present study were approved by the Animal Investigation Ethics Committee of Nanjing Medical University (2,003,001-1), following the institutional guidelines and national animal welfare.
After a week of acclimation, a total of forty mice (20-23 g, 2 months old) were randomly divided into four groups (N = 10 for each group) and orally administered butyl benzyl phthalate (BBP, Sigma-Aldrich, St. Louis, MO, USA) at doses of 5, 15, and 50 mg/ kg/day by gavage for 2 months. The control group was treated with the same volume of vehicle (corn oil). The exposure doses in this study were designed according to the individual tolerable daily intakes (TDI) (500 μg/kg/day) imposed by the European Food Safety Authority (EFSA) (EFSA 2005). The animal equivalent dose was calculated based on the formula of the body surface area normalization method (Reagan-Shaw et al. 2008). Body weight was quantified every 3 days. One day after the last treatment, four mice from each group were sacrificed by non-anesthetic cervical dislocation to avoid potential deleterious bias since anesthetic might affect the nervous system. At the same time point, the remaining six mice from each group were subjected to the Morris water maze (MWM) test, and 1 day after the behavioral test, the mice were sacrificed by cervical dislocation. The cerebral tissues were immediately isolated and collected for subsequent analysis.

MWM test
A standard 5-day MWM test was performed to evaluate the spatial learning and cognitive function of experimental mice . The MWM setup contained a water maze pool (90 cm in diameter) and an escape platform (9 cm in diameter). The water maze pool was filled with opaque water, while the platform was fixed in the center of one of the two hemispheres. In the place navigation task (days 1-4), mice were trained to find the hidden escape platform in four different starting positions (intersession interval 30-60 min), starting at 9:00 AM, per day. The monitoring time for each mouse swim was set to 120 s, and mice that failed to find the target platform within 120 s were manually guided to the platform. All mice were gently released into the water and allowed to stay on the escape platform for 10 s. In the spatial probe test (day 5), the platform was removed, and the starting position was set at the opposite hemisphere of the platform hemisphere. Swim speed, the latency of first time entering the platform area, and the frequencies of crossing the platform area were recorded using a camera system and analyzed with Noldus Ethovision XT software (Wageningen, Netherlands).

Histopathological analysis
After euthanasia, the freshly isolated brain (N = 4 for each group) was fixed with 4% paraformaldehyde, followed by routine dehydration, paraffin embedding, and serial slicing (5 μm). The cerebral sections were subsequently either stained with hematoxylin and eosin (H&E) or Nissl staining solution, following the corresponding manufacturer's instructions (Solarbio, Beijing, China). The numbers of surviving neurons with intact cell morphology were counted in per 100μm length of the hippocampal CA1, CA3, and DG regions. Images were captured and analyzed using a Pannoramic MIDI scanner (3DHISTECH, Budapest, Hungary).

Transcriptomics
The hippocampal samples (N = 3 for each group) from 50 mg/kg/day BBP-treated mice and control mice were freshly isolated and sent to LC-Bio Technology (Hangzhou, Zhejiang, China) for RNA sequencing and bioinformatic analysis. Briefly, total RNA was extracted using TRIzol reagent (Invitrogen, CA, USA) and subjected to 2 × 150 bp (PE150) paired-end sequencing with an Illumina Novaseq™ 6000 in accordance with the vendor's recommended protocol. For each sample, HISAT2 software was applied to assemble mapped reads and the sequence results were calculated as fragments per kilobase of exons per million reads (FPKM). The differentially expressed genes were selected with Log 2 (fold-change) > 1.5 or < − 1.5 and P value < 0.05. Gene Ontology (GO; http:// geneo ntolo gy. org/) and Kyoto Encyclopedia of Genes and Genomes (KEGG; https:// www. genome. jp/ kegg/) analyses were then used to identify the functional and signaling pathways of the screened differentially expressed genes.

Quantitative real-time PCR (qRT-PCR) analysis
Total RNA of the hippocampus samples (N = 3 for each group) was isolated using RNA isolater Total RNA Extraction Reagent (Vazyme, Nanjing, China) in accordance with the manufacturer's instructions. Complementary DNA (cDNA) was synthesized with 1 μg total RNA using HiScript II Q Select RT SuperMix (Vazyme, Nanjing, China). Realtime PCR was performed with routine activating (95 °C, 5 min), denaturing (95 °C, 15 s), and annealing (72 °C, 1 min) programs to evaluate the expression of mRNA. Specific gene primers were designed by Generay and are presented in Table S11.

Statistical analysis of laboratory data
The animal results were expressed as the mean ± standard deviation (SD) and analyzed with SPSS 17.0 (SPSS Institute, Chicago, IL, USA). Statistical differences in frequencies in mice crossing the target site between control and BBP treatment groups were analyzed using the Kruskal-Wallis independent nonparametric test, whereas differences in other outcomes were analyzed using one-way analysis of variance (ANOVA) test, followed by Student-Newman-Keuls multiple comparison test. A value P ≤ 0.05 was considered statistically significant.

Demographic characteristics of the study individuals
In this study, as illustrated in Table 1, a total of 836 elderly participants were identified from NHANES, 2011-2014. Among them, the ethnicity and educational levels of most individuals were non-Hispanic white 46.7% and some college or AA degree 28.8%, respectively. The population 60 to 65 years old was the largest population (38.0%), and 67.7% participants had no or low physical activity. Approximately 16.5%, 36.2%, 20.7%, and 6.5% of elders were recognized to have the conditions of poverty, diabetes, CVD, and stroke, respectively. More than 70% of participants had the condition of hypertension, overweight/obesity (BMI ≥ 25), and alcohol use. The proportion of gender and smoking status were found to be approximately "fifty to fifty." The median concentration of urinary creatinine was 98.0 mg/100 mL. The results of Pearson correlation analysis (Fig. S2) demonstrated that most phthalate metabolites were significantly correlated among themselves, except for MEP, which was uncorrelated with any of the other metabolites. MCOP was strongly correlated with MCNP (r = 0.71, P < 0.001) and MCPP (r = 0.75, P < 0.001). In addition, there was a strong and positive correlation between MiBP and MBP (r = 0.80, P < 0.001). As expected, the four metabolites of DEHP (MEHP, MECPP, MEOHP, and MEHHP) were strongly correlated with each other, suggesting potential co-exposure patterns of these phthalate metabolites.

Independent association between phthalates and cognitive impairment
Survey-weight logistic regression was conducted to estimate the independent relationship of phthalate metabolites and cognitive impairment in elders, and to better show the independent associations, phthalates were further grouped into HMW (Fig. S3) and LMW (Fig. 1) phthalates. In total, MBzP was positively related to AF (P trend = 0.009) and DSST (P trend = 0.042) impairment. Comparing the highest exposure quantile with Q1, the OR (95% CIs) for MBzP with AF was 2.89 (1.36 to 6.15), while the ORs (95% CIs) for MBzP with IR and DR impairment were 2.53 (1.31 to 4.88) and 2.26 (1.06 to 4.25), respectively, in Q3 versus (vs.) Q1. MBzP was also positively linked to DSST impairment in Q2 vs. Q1 (OR: 2.45, 95% CI 1.05 to 5.89) and Q4 vs. Q1 (OR: 2.82, 95% CI 1.02 to 7.82). For other metabolites, a significantly positive association was only observed in MEHP with DSST impairment (OR: 1.96, 95% CI 1.05 to 3.64, Q4 vs. Q1) and other associations were either nonsignificant or negatively significant: MEOHP was inversely related to AF impairment in Q2 vs. Q1 (OR: 0.41, 95% CI: 0.20 to 0.82) and MCNP was negatively associated with AF impairment in Q2 vs. Q1 (OR: 0.52, 95% CI: 0.29 to 0.93). All the statistically significant and positive results remained stable in the unweighted sensitivity analysis (Table S2), except for the associations of MBzP with DR and DSST impairment, which became nonsignificant.

Combined association between phthalates and cognitive impairment
PCA was conducted to further evaluate the combined association of phthalates and cognitive deficits. Our data were suitable for PCA, as the KMO value was 0.89 and a significant correlation among phthalates was identified by Bartlett's test (P < 0.05). As shown in Fig. S4, we identified three PCs with eigenvalues > 1 for subsequent analysis, and the cumulative variance percentage was 77%, with variance percentages of 33%, 23%, and 21% for PC1, PC2, and PC3, separately. Figure S5 and Fig. S6 illustrate the score plot and loading plot of our PCA respectively. The selected PCs were named according to the top 3 contributions and prior knowledge, and as shown in Table S3 Combined associations between principal components and cognitive impairment are also shown in Fig. 1 and Fig. S3: LWM phthalates (PC3) were positively associated with DSST impairment (P trend = 0.022), with an OR (95% CIs) of 1.93 (1.01 to 3.69) in Q4 vs. Q1, whereas the association of LWM phthalates with DR impairment was noted to be negative in Q2 vs. Q1 (OR: 0.54, 95% CI: 0.31 to 0.94). In addition, non-DEHP HWM phthalates (PC2) were negatively linked to DSST impairment (P trend = 0.009), and comparing the highest exposure quantile with Q1, the corresponding OR (95% CIs) was 0.49 (0.26 to 0.93) (Fig. S3). No significant association was observed for DEHP HWM phthalates with any cognitive test impairment (Fig. S3).
Next, the WQSR was applied to validate the combined impact of phthalates on cognitive impairment on the basis of PCA results. Table S4, Table S5, and Table S6 show the weights of phthalates in non-DEHP HWM or LWM phthalates in the models of the WQS index with DSST or DR impairment. In general, MBzP was the main contributor when the direction was positive, while for the negative direction, MiBP weighted the most. Figure 2A demonstrates that a significantly positive association was noted in DSST impairment with the WQS index containing top the 7 phthalates and 8 phthalates in the LWM phthalates component (OR 7-phthalates : 1.56 95% CI: 1.03 to 2.36, OR 8-phthalates : 1.55 95% CI: 1.02 to 2.35). Moreover, MBzP had the highest weight in the association of both 7-phthalates and 8-phthalates WQSR model, with the weight being 40.1% and 38.7%, respectively (Fig. 2B). The non-linear association between WQS index containing the top 7 phthalates and 8 phthalates in the LWM phthalates component and incidence of DSST impairment is shown in Fig. 2C and although no apparent violation of linearity was observed (P > 0.05), the ORs of DSST impairment were significantly increased in cases with higher WQS index (P-non-linear was 0.8685 and 0.7984 for 7 and 8 phthalates, respectively). However, the associations of LWM phthalates with DR impairment (Table S7) and non-DEHP HWM phthalates with DSST impairment (Table S8) became nonsignificant in the WQSR model. In the positive direction of the all-phthalates WQSR model, the associations between WQS index and the four cognitive test impairment were nonsignificant (Table S9). correspond to impairment of IR, DR, AF, and DSST, respectively. The model was adjusted for age, sex, race, PIR, education level, BMI, physical activity, smoking status, drinking status, stroke, diabetes, hypertension, CVD, and urinary creatinine. Individuals who scored < 17 for IR, < 5 for DR, < 14 for AF, or < 34 for DSST were considered to have potential cognitive impairment. LWM, low molecular weight; IR, immediate recall; DR, delayed recall; AF, animal fluency; DSST, Digit Symbol Substitution Test; CVD, cardiovascular disease. The red box represents a significantly positive association while green box represents a significantly negative association. *P < 0.05, **P < 0.01 Nevertheless, MBzP exhibited the highest contribution for all outcomes (Fig. 3).

BBP-induced learning and memory impairment in mice
As described above, MBzP is a major metabolite of BBP and was the only derivative we found with an obvious positive association with cognitive impairment in elders. We therefore chose to use BBP instead of MBzP in our animal experiments to further validate the detrimental effects of phthalates on cognitive functions. Laboratory mice were orally administered a series dose of BBP (0, 5, 15, and 50 mg/kg/day) for 2 months. During the whole study period, as shown in Fig. S7, the body weight and brain organ coefficient of mice were not significantly altered by BBP treatment, indicating that no obvious side effects were induced by BBP treatment.
Next, we estimated the effects of BBP on mouse neurobehavioral function using the MWM test. Among the four groups, the average swimming speed of mice revealed no significant difference, suggesting an even motor ability (Fig. 4A). In the spatial learning trials, the escape latency of BBP-treated mice was significantly increased compared that of control mice on the second day (67.32 ± 16.30 s, P < 0.05 in the 5 mg/kg BBP group) and fourth day (37.58 ± 7.82 s, P < 0.05 in the 50 mg/kg BBP group) of training (Fig. 4B). In the probe trials, BBP-treated mice spent significantly more time finding the target site (55.61 ± 23.47 s, P < 0.01 in the 50 mg/kg BBP group) (Fig. 4C). In addition, a significant dose-dependent decreasing trend in the frequencies of mice crossing the target site was observed in the BBP-treated groups (P < 0.05) (Fig. 4D). Representative images of mice motion trajectories (Fig. 4E) showed that compared with the control mice, BBP-treated mice spent more time in the areas surrounding the water maze pool and the signs of searching for the platform were significantly decreased. These findings collectively illustrate that BBP disrupts cognitive functions in mice.

BBP-induced disturbance of hippocampal neurons in mice
Next, we examined the effect of BBP on hippocampal neurons, as the hippocampus serves a purely cognitive structure involved in memory and navigation. As illustrated in Fig. S8 and Fig. 5A, HE and Nissl staining of brain sections showed the normal morphology of neurons with orderly cell arrangement in the CA1, CA3, and DG regions of control mice. In contrast, a significant increase in nuclear pyknosis was observed in the DG region of mice treated with BBP at doses of 15 mg/kg/day and higher (Fig. S8). In addition, neurons in the CA1, CA3, and DG regions were found to be loosely arranged, lightly stained, swollen, and ruptured in the highest dose group (Fig. 5B). The counts of surviving neurons Fig. 2 Combined association between the WQS index containing the top phthalates in LWM phthalate mixtures and DSST impairment in WQSR analysis. A Combined association between the WQS index containing the top phthalates in LWM phthalate mixtures and DSST impairment in WQSR analysis. B WQS weights in the WQSR model between DSST impairment and WQS index of LWM phthalate mixtures. C Restricted spline curves for association between phthalate mixtures (WQS index) and DSST impairment. The direction was assumed to be positive based on the results of principal components analysis. The model was adjusted for age, sex, race, PIR, education level, BMI, physical activity, smoking status, drinking status, stroke, diabetes, hypertension, CVD, and urinary creatinine. Individuals who scored < 34 for DSST were considered to have potential cognitive impairment. WQSR, weighted quantile sum regression; DSST, Digit Symbol Substitution Test; CVD, cardiovascular disease; LWM, low molecular weight. The red box represents a significantly positive association. *P < 0.05 in the CA1 and DG regions were significantly decreased in the 50 mg/kg/day BBP administration group ( Fig. 5C and E) compared with the control group, while Fig. 5D shows that BBP slightly reduced the counts of surviving neurons of CA3 region in the 5 mg/kg/day BBP administration group. These results indicate that BBP impairs the structure of hippocampus and specifically disrupts the hippocampal neurons.
To gain insight into the effect of BBP on the global gene expression of the hippocampus, transcriptomic analyses of the hippocampus in mice treated with corn oil vehicle or 50 mg/kg/day BBP were performed. Three independent hippocampal replicates from each group were analyzed to evaluate mRNA changes, and the filtering criteria were set as P < 0.05 and expression fold change > 1.5. In comparison with the control mice, we identified a total of 431 differentially expressed genes in BBP-treated mice (259 upregulated and 172 downregulated), which was visually displayed by volcano plot (Fig. 6A). GO functional analysis (Fig. 6B) showed that, for both downregulated and upregulated genes, the majority were enriched in biological process, regulation of transcription, and signal transduction in biological processes, while most cellar components were enriched in membrane, integral component of membrane, and plasma membrane. The top 3 enriched pathways for molecular function were protein binding, metal ion binding, and molecular function. Unlike the pattern observed in GO analysis, the results of KEGG enrichment analysis revealed remarkable differences between upregulated and downregulated genes. As demonstrated in Fig. 6C, the upregulated mRNAs were mostly enriched in the Fig. 3 Weights of phthalates between the WQS index of the whole phthalates and cognitive impairment. Annotations (A), (B), (C), and (D) correspond to impairment of IR, DR, AF, and DSST, respectively. The direction was assumed to be positive and the model was adjusted for age, sex, race, PIR, education level, BMI, physical activ-ity, smoking status, drinking status, stroke, diabetes, hypertension, CVD, and urinary creatinine. Individuals who scored < 17 for IR, < 5 for DR, < 14 for AF, or < 34 for DSST were considered to have potential cognitive impairment. WQSR, weighted quantile sum regression; DSST, Digit Symbol Substitution Test; CVD, cardiovascular disease cytokine-cytokine receptor interaction pathway, while the majority of downregulated mRNAs belonged to the neuroactive ligand-receptor interaction pathway. Many genes screened from the above pathways are critical for the regulation of cerebral inflammation (e.g., tnf, il11, and il12a) or neurotransmitters (e.g., htr2c, ptger1, and chrna3). We therefore selected them as candidates for the validation of mRNA expression using quantitative realtime PCR (qRT-PCR) analysis. As shown in Fig. 6D, all the qRT-PCR results remained consistent with the trend observed in the transcriptome analysis. Taken together, the above data authenticate the deleterious effect of BBP on hippocampal neurons, suggesting that BBP-induced cognitive impairment in mice could be partially caused by disrupted hippocampal neurons and altered expression of neuroactive receptors and cytokine-related genes.

Discussion
In the present study, we found an independent and positive association between urinary MBzP, the major metabolite of BBP, and cognitive impairment in elderly NHANES subjects, while the association was nonsignificant or negative-significant for other phthalate metabolites. More importantly, the results of two-stage PCA-WQSR combined association analysis revealed that LWM phthalates were linked to cognitive dysfunctions and that MBzP contributed most to this potential cognitive detrimental effect. Considering the relatively low levels of urinary MBzP among all phthalates, these findings collectively suggested the critical role of MBzP/BBP in the process of cognitive impairment. In addition, BBP decreased learning Fig. 4 BBP-induced learning and memory impairment in mice. Laboratory mice (2 months old) were administered doses of 5, 15, and 50 mg/kg/day BBP or the same volume of corn oil by gavage for 2 months. At the study endpoint, the Morris water maze (MWM) test was performed for 6 mice of each group. A Across the whole MWM test, no significant difference was observed in the swimming speed of mice from the four groups. B In the navigation trial, BBP-treated mice spent significantly more time finding the platform than control mice. C In the probe trial, the escape latency to first target crosso-ver of BBP-treated mice was significantly higher than that of control mice. D In the probe trial, compared with corn oil treatment, BBP treatment significantly decreased the frequencies crossing the platform zone. E Representative images of typical swimming traces in the spatial memory test. The data are expressed as the mean ± SD of 6 mice per group. One-way ANOVA or Kruskal-Wallis independent nonparametric test was used and *P < 0.05, ** P < 0.01 compared with the control group and memory functions by impairing hippocampal neurons and especially disrupting the neuroactive ligand-receptor interaction pathway in mice. To the best of our knowledge, we are the first to provide population and confirmatory experimental evidence that cognitive function exhibits susceptibility to BBP toxicity and that the disruption of hippocampal neurons and the inhibition of synaptic transmission may be the underlying mechanism for this.
Elderly individuals are a particularly vulnerable population among whom neurodegenerative diseases, including cognitive impairment largely develop (Hou et al. 2019). This physiological declining of cognitive function also made elder adults a more sensitive population for environmental toxicants (Lee et al. 2022). Indeed, as a nonnegligible and modifiable risk factor for cognitive impairment, environmental pollution has raised dramatic public awareness and its implication for the prevention of neurodegenerative disorders has been extensively explored (Marras et al. 2019;Periñán et al. 2022). In this study, positive pattens of phthalates with cognitive impairment remained statistically significant after the models adjusting for age, suggesting that phthalate exposure may serve as an independent risk factor for cognitive deficits in elderly population. Nevertheless, evidence regarding the association between phthalate and cognitive impairment in elders is rather insufficient. In a study utilizing the same database as the present studydata from NHANES 2011-2014 cycles (Weng et al. 2022) -the authors found inconsistent and confusing results for different tests and methods: MECPP, MnBP, MCOP, and MCPP were negatively linked to IR z scores, while MCPP and MBzP were negatively linked to DR z scores. Additionally, the main effect of phthalate mixture was remarkably altered between the WQSR and Bayesian kernel machine regression (BKMR) models. In contrast, our findings of Data are expressed as the mean ± SD of 3 mice per group. One-way ANOVA was used and *P < 0.05, **P < 0.01 compared with the control group single phthalate based logistic regression analysis and mixed phthalate based two-stage PCA-WQSR analysis collectively revealed a consistent and positive association of the same phthalate, MBzP, from LMW phthalates (PC3), with cognitive impairment in elderly individuals. The discrepancy between Weng's work and ours might be that we included several variables, such as CVD status and MEHP, which were not considered in Weng's study. In addition, we set a cutoff value for each test to distinguish individuals with potential cognitive impairment from healthy individuals and designed a two-stage PCA-WQSR analysis that divided the phthalate metabolites into different PCs and estimated the co-exposure correlation and contribution more robustly.
Our epidemical data showed that MBzP was the only metabolite that was positively associated with the four cognitive outcome impairments in single pollutant models, and the co-exposure pattern remained significantly positive in DSST deficit with LMW phthalates, in which MBzP was the predominant contributor. These findings were partly in accordance with a previous epidemiologic study respecting the impact of phthalates on children's cognitive functions, which reported an inverse association of MBzP with children's full-scale IQ (Li et al. 2019). Moreover, as shown in Table S2, the concentrations of MBzP in the urinary samples were relatively low among all the phthalates, further implying the vulnerability of cognitive functions to MBzP toxicity.
Apart from the increased cognitive deficit risk observed for MBzP, we also found a positive association of MEHP, a metabolite of DEHP, with IR impairment in Q4 vs. Q1 of a single phthalate analytic model. Nonetheless, the association became statistically nonsignificant in co-exposure of DEHP HWM phthalates (PC1) and WQSR models. Considering the biased estimation and inflated false caused by the collinearity of phthalates in single pollutant analysis (Pitard and Viel 1997), the finding that MEHP was associated with cognitive impairment in elders should be interpreted with caution. However, as a typical representative of phthalates, DEHP was reported to disrupt spatial learning and memory abilities, impair blood-brain barriers and decrease synaptic proteins in experimental animals (Ahmadpour et al. 2021;Sun et al. 2021). One possible explanation for this observation is that the concentrations of DEHP used in those experimental animals were much higher than the environmental DEHP exposure of humans in this study. In addition, we also observed negative associations of MEOHP and MNCP with AF impairment in Q2 vs. Q1 of the single pollutant model and non-DEHP HWM phthalates (PC2) with DSST impairment in Q4 vs. Q1, and with DR impairment in Q2 vs. Q1 of the PCA co-exposure model. These results should not be simply deduced as protective factors as the simple design of single pollutant model might cause potential bias, and moreover, although statistically nonsignificant, the ORs Fig. 6 BBP-induced transcriptomic alteration of hippocampal genes in mice. Laboratory mice (2 months old) were administered doses of 5, 15, and 50 mg/kg/day BBP or the same volume of corn oil by gavage for 2 months. At the study endpoint, hippocampal samples were collected. A Volcano plots of differentially expressed genes in the 50 mg/kg/day BBP treatment group versus the control group. Blue dots indicate the downregulated genes, while red dots show the upregulated genes in the BBP administration group. The differ-entially expressed genes were selected with the criteria of P < 0.05 and Log 2 |FC|> 1.5. B GO gene set enrichment analysis of differentially expressed genes. C KEGG pathway enrichment analysis of differentially expressed genes. D Validation of candidate differentially expressed genes with qRT-PCR analysis. E Results were summarized as graphic illustrations. Data are expressed as the mean ± SD of 3 mice per group. One-way ANOVA was used and *P < 0.05, **P < 0.01 compared with the control group were > 1 in Q4 vs. Q1. Besides, the results of the PCA coexposure model became nonsignificant in the WQSR validation analysis.
It is acknowledged that urinary phthalate metabolites can serve as proper internal exposure biomarkers and are suitable surrogates of parent prototypes (Reeves et al. 2021). Therefore, as mentioned above, BBP was selected as the potential cognitive-toxic candidate for the subsequent causality analysis in this study. To highlight the role of BBP in the progression of cognitive impairment, mouse of 2 months was used for the BBP exposure experiments as it is a common and cognition-sensitive animal model used for assessment of cognitive functions by previous studies (Fröhlich et al. 2016;Ni et al. 2021). As a nonnegligible member of the LWM phthalate family, BBP was considered a priority pollutant by the European Commission and US Environmental Protection Agency (Silano et al. 2019;USEPA 2014). The neurotoxicity of BBP has also gained public attention. For instance, a systematic review of evidence from birth cohorts reported that prenatal exposure to BBP was linked to decreased cognitive scores and behavior in children (Zhang et al. 2019). Moreover, given the lower water solubility of BBP than other phthalates (Xie et al. 2011), which is an important factor that is inversely associated with the neurotoxicity of phthalates (Screnci et al. 2000), it is theoretically plausible that BBP may serve as a critical neurotoxic phthalate. However, to date, regarding cognitive impairment, only one experimental study has been conducted to explore the effect of BBP on cognitive function in mice, and their findings should be interpreted with caution because the doses they applied were relatively high, which were 10 to 25 times higher than ours (Min et al. 2014). In this study, we found that BBP exposure resulted in pathological disruption of the hippocampus and a reduction in learning and memory abilities in mice at relatively low doses of BBP (50 mg/kg/day), which was selected to mimic the total tolerable intake of BBP in mice, further validating the essential role of BBP as a potential neurotoxic phthalate in the progression of cognitive impairment.
Hippocampal genes are considered "key modulators" in cognitive functions including learning and memory (Li et al. 2022). To understand the potential mechanism behind the detrimental effect of BBP on cognition, which has rarely been explored previously, the expression profile of genes in the hippocampus was analyzed in control and 50 mg/ kg/day BBP treated mice. Our findings of KEGG pathway analysis demonstrated that the downregulated DEGs were mainly related to the neuroactive ligand-receptor interaction pathway, a group of membrane receptors involved in the identification of neuroactive ligands such as the classic neurotransmitters serotonin, acetylcholine, and dopamine, which modulate learning and memory directly (Huang et al. 2016). Moreover, GO annotation results further confirmed that the top enriched variational expression genes were related to the modulation of membrane and protein binding processes, and subsequently affected intercellular neural signal transduction, recognition, and transport (Vit and Petrak 2017). The decreasing trend of neuroactive receptors htr2c, ptger1, and chrna3 was also validated in our qPCR analysis, which were reported to identify serotonin, amplify dopamine receptors, and modulate nicotinic acetylcholine, respectively (Hebras et al. 2020;Schuch et al. 2016;Tanaka et al. 2009). Thus, the disrupted functions of these DEGs may be the potential mechanism by which BBP induced cognitive decline in mice. In addition, as illustrated in Fig. 6C, most of the upregulated DEGs were enriched in cytokine-cytokine receptor interaction pathway, which regulates neuroinflammation through cytokines like TNF, IL11, and LIL12a . TNF signaling, by interacting with the corresponding specific receptors, modulates the strength of synaptic transmission and directly inhibits synaptic plasticity in the hippocampus (Liu et al. 2017). IL11 is a member of IL6-type cytokine family that is involved in the process of acute-phase response to infection or environmental injury, and it was found to be increased in the cerebrospinal fluid (CSF) of patients with cognitive decline (Galimberti et al. 2008). The excitement of IL12a was associated with proinflammatory reaction and contributed to reduced performance in the processing speed of elders (Lin et al. 2019). Therefore, in combination with cognitive and historical results, we deduced that BBP interfered with cognitive function by activating hippocampal inflammation and subsequently disrupting neuroactive receptors.
This research possesses several advantages. First, the epidemical data were representative of subjects recruited from the national population-based program NHANES. Second, the survey weighted logistic single pollutant model and twostage PCA-WQSR mixed pollutant model were collectively designed to identify the potential candidate with cognitive toxicity, which estimates the joint association more robustly and evaluates the contribution more intuitively. Third, as the cross-sectional design of this study may be limited by the incapability to conclude a causal relationship, experimental mice were further used to confirm the possible causality of BBP with cognitive impairment and to explore the potential mechanisms, which made it more reliable to interpret the results.

Conclusion
Taken together, by using survey weight logistic-based singlepollutant and PCA-WQSR-based mixed-pollutant analyses, our research indicated that MBzP, the major metabolite of BBP, was the predominant contributor to the increased cognitive impairment of the elderly population. Further confirmatory experiments revealed that the disrupted hippocampal neurons, especially the decreased neuroactive receptors and activated inflammation induced by BBP, are a potential mechanism of interference. These findings identify, emphasize, and validate the critical role of BBP from phthalate mixtures, which may provide new insight into phthalate-induced cognitive toxicity in elderly individuals and help reveal the epidemiological and toxicological importance of BBP-related cognitive deficits. Nonetheless, as the phthalate metabolites were evaluated based on a single measurement of spot urine sample from each study subject, exposure misclassification may exist due to the short half-life of phthalates (Braun et al. 2012) and the correlation between phthalates and cognitive deficits should be carefully concluded. Although the elderly subjects maintain a consistent use habit of plastic materials, and resultantly the single spot urine may also represent the long-term exposure level of phthalates to a certain extent (Hoppin et al. 2002;Liu et al. 2022), further cohort studies with long-term phthalates exposure data are required to disclose the temporal association between cognitive deficits and phthalates.
Author contribution Yongquan Yu and Shou-Lin Wang conceptualized the study and contributed to the overall organizing of the experiments; Yucheng Wang, Shuge Shu and Di Zhang conducted the data analysis; Yu Dong, Jiayi Xu, Ying Zhang, and Wei Shi performed the experiments. All authors read and approved the final manuscript.
Funding This work was supported by the National Natural Science Foundation of China (grant numbers 82003494, 81973091, 82173562), ZhiShan Scholar Program of Southeast University (grant numbers 2242022R40061), and China Postdoctoral Science Foundation (grant numbers 2020M681672).

Declarations
Ethics approval and consent to participate This study uses secondary epidemical data analyses without any personal information identified using statistical data from the NHANES website and therefore did not require review. The laboratory procedures in the present study were approved by the Animal Investigation Ethics Committee of Nanjing Medical University.

Competing interests
The authors declare no competing interests.