N⁶-Methyladenosine (M⁶A) in Fetal Offspring Modifies Mitochondrial Gene Expression following Gestational Nano-TiO₂ Inhalation Exposure

Abstract

Background:

N⁶-methyladenosine (m⁶A) is the most prominent epitranscriptomic modification to RNA in eukaryotes, but it’s role in adaptive changes within the gestational environment are poorly understood. We propose that gestational exposure to nano titanium dioxide (TiO₂) contributes to cardiac m⁶A methylation in fetal offspring and influences mitochondrial gene expression.

Methods:

10-week-old pregnant female FVB/NJ wild-type mice underwent 6 non-consecutive days of whole-body inhalation exposure beginning on gestational day (GD) 5. Mice were exposed to filtered room air or nano-TiO₂ with a target aerosol mass concentration of 12 mg/m³. At GD 15 mice were humanely killed and cardiac RNA and mitochondrial proteins extracted. Immunoprecipitation with m⁶A antibodies was performed followed by sequencing of immunoprecipitant (m⁶A) and input (mRNA) on the Illumina NextSeq 2000. Protein extraction, preparation, and LC-MS/MS were used for mitochondrial protein quantification.

Results:

There were no differences in maternal or fetal pup weights, number of pups, or pup heart weights between exposure and control groups. Transcriptomic sequencing revealed 3,648 differentially expressed mRNA in nano-TiO₂ exposed mice (Padj≤0.05). Transcripts involved in mitochondrial bioenergetics were significantly downregulated (83 of 85 genes). 921 transcripts revealed significant m⁶A methylation sites (Padj≤0.10). 311 of the 921 mRNA were identified to have both 1) significantly altered expression and 2) differentially methylated sites. Mitochondrial proteomics revealed decreased expression of ATP Synthase subunits in the exposed group (P≤0.05).

Conclusions:

The lack of m⁶A modifications to mitochondrial transcripts suggests a mechanism for decreased transcript stability and reduced protein expression due to gestational nano-TiO₂ inhalation exposure.

Graphical Abstract

Introduction

Epigenetic reprogramming is one of the main mechanisms responsible for adverse effects on long-term health outcomes following environmental exposure to toxicants (Barouki et al. 2018; Trevino et al. 2020). While studies of transient DNA and histone modifications have provided significant insight into the field of toxicology, the observation of chemical modifications at the level of mRNA has introduced a new layer of complexity. The study of mRNA chemical modifications, otherwise known as epitranscriptomics, provides a new perspective for understanding how alterations in the stability of mRNA transcripts can impact protein expression in the maternal environment and shape the health of the offspring. N⁶-methyladenosine (m⁶A) is the most abundant mRNA modification in eukaryotes (Zaccara et al. 2019) and has the ability to alter the translational capacity of transcripts. There is currently very little information regarding how m⁶A modifications impact fetal development following gestational particulate or engineered nanomaterial (ENM) exposure.

The epigenetic modifications implicated in the long-term effects following environmental exposure thus far, have centered primarily on oxidative stress, DNA methylation, and histone post-translational modifications (Ruiz-Hernandez et al. 2015; Seo et al. 2020). However post-transcriptional modifications and their role in regulating RNA function have recently gained traction. Methylation of bases can alter RNA structure, folding, stability, degradation, localization, and thereby its interactions with other RNAs or proteins (Kumar and Mohapatra 2021; Zaccara et al. 2019). Recently, m⁶A has been suggested to play a prominent role in epitranscriptomic reprogramming following particle inhalation exposure (Cayir et al. 2019; Cui et al. 2021; Kunovac et al. 2021), occurring in ~30% of transcripts and playing a pivotal role in numerous physiological systems (Yang et al. 2020). In addition to being highly abundant in mammalian mRNA, it is also vastly conserved across eukaryotic species including, yeast, plants, and mammals (Dominissini et al. 2012; Fu et al. 2014). Dysregulation of m⁶A can alter gene expression, which can lead to changes in cellular function and the onset of cancer, psychiatric disorders, metabolic derangements, and cardiovascular disease (Yang et al. 2020; Zaccara et al. 2019).

A predominant cause of cellular dysfunction following gestational exposure is mitochondrial maladaptation. Gestational exposure to a pernicious environment, including high glucose (Kim et al. 2015), maternal obesity (Baker et al. 2017; Boyle et al. 2017), and air pollution (Gruzieva et al. 2017), can impact epigenetic reprograming in fetal mitochondrion leading to adverse outcomes. Following gestational ENM inhalation exposure, changes in DNA methylation (Kunovac et al. 2019) and epitranscriptomic modifications of reactive oxygen species (ROS) scavenging transcripts (Kunovac et al. 2021) have been associated with higher ROS levels and decreased mitochondrial electron transport chain complex activities. Ultimately, gestational nano-TiO₂ inhalation exposure can result in functional, cardiac deficits, including reduced ejection fraction and fractional shortening in adult animals as well as subtle changes in left ventricular strain at the fetal stage (Kunovac et al. 2019). Additionally, enhancement of ROS scavenging in fetal progeny exposed during gestation can ameliorate deficits seen at the adult stage, specifically in normalizing ejection fraction and fractional shortening measurements (Kunovac et al. 2021).

The overall landscape of m⁶A, and its contribution to disease in the heart, remains to be elucidated in fetal offspring following gestational ENM inhalation exposure. In this study, we implement transcriptome-wide mapping of m⁶A by combining RNA immunoprecipitation and RNA sequencing (m⁶A-RIP-seq) to clarify the role of m⁶A methylation on diminished cardiac and mitochondrial function in fetal offspring following gestational exposure to nano-TiO₂ (Kunovac et al. 2021; Kunovac et al. 2019). Lack of m⁶A methylation to mitochondrial transcripts likely decreases the stability of those transcripts and results in decreased protein expression. Our data emphasize how gestational exposure to ENMs can alter m⁶A methylation and can affect mitochondrial bioenergetics.

Materials and Methods

Animal Model

Friend Virus B NIH Jackson (FVB/NJ) wild-type mice (14 females, 7 males at 10 weeks) were purchased from Jackson Laboratory (Bar Harbor, ME). Animals were housed in the West Virginia University Health Sciences Center Animal Facility with access to a standard chow diet and water ad libitum. All animals were allowed a 48-hour acclimation period before being handled. Females were housed in groups of 4 per cage and introduced to male bedding three days prior to being introduced to the males for breeding. This was implemented to take advantage of the Whitten effect, which states that females are most receptive to mating on the third night after exposure to male pheromones that are found in urine (Whitten 1956). This allowed us to synchronize the females’ estrous cycles prior to mating and therefore increased the number of females becoming pregnant at similar time points, leading to well age-matched cohorts.

A harem breeding strategy was utilized with two females and one male per cage set up on the third night following exposure to male bedding. Females were checked for vaginal plugs in the early morning after they were set-up with the males. Pregnancy was verified with identification of a vaginal plug at which point females were randomly separated into sham (7 pregnant dams) and nano-TiO₂ (7 pregnant dams) groups. On approximately gestational day (GD) 5, maternal inhalation exposure was initiated for each respective exposure group (sham or nano-TiO₂) that occurred for 6 hours/day, for 6 days over an 8-day period. Sham and nano-TiO₂ exposed females were treated identically with the exception of the inhaled particulate (i.e., diet, housing, acclimatization, etc.). Pregnant dams (7 sham, 7 nano-TiO₂) were humanely killed on GD 15 and the pups were removed from the uteri; GD 15 marks the end of cardiac organogenesis, which is the first time point in which all the cardiac structures have formed/matured, including separation of the aorta and the pulmonary artery (GD 13) and ventricular and atrial septation (GD 14.5) (Keller et al. 1996; Vuillemin and Pexieder 1989). Dams were humanely killed using 5% isoflurane followed by cervical dislocation. The fetal pups were subsequently harvested and tissue collected.

The dams, pups, and pup hearts were weighed, and weights recorded (Table 1). The hearts of fetal offspring were pooled per each litter which was considered n=1. Due to limited tissue (~0.003 grams wet weight per heart), pooling of fetal hearts was necessary to achieve adequate protein and mRNA concentrations for downstream applications. Therefore, it was not possible to identify sex differences at either the genomic or proteomic level in the current study, as tissue was pooled irrespective of sex. The hearts of offspring of 3 sham and 3 nano-TiO₂-exposed dams were saved in RNAprotect Tissue Reagent (Item no. 76105; Qiagen, Hilden, Germany) in cryotubes, flash frozen, and stored in −80°C for future analysis. The hearts of offspring of the other 4 sham and 4 nano-TiO₂-exposed dams were immediately processed for mitochondrial isolation as described below. The West Virginia University Animal Care and Use Committee approved all animal studies which conformed to the most current National Institutes of Health (NIH) Guidelines for the Care and Use of Laboratory Animals (8^th edition) manual.

Table 1: Maternal and fetal pup characteristics.

Dams from sham (n=5) and nano-TiO₂ (n=5) inhalation exposure were assessed. All weights given are wet weights. A 2-sided Student’s t-test was used for two normally distributed, continuous variables. The Shapiro-Wilk test was utilized for normality testing. For non-Gaussian distributed data, the Mann-Whitney U test was applied.

Sample	Maternal Weight (g)	Number of Pups in Litter	Average Pup Weight (g)	Average Pup Heart Weight (g)
Sham Exposure
1	31.9	8	0.61	0.0038
2	34.1	10	0.56	0.0030
3	34.1	10	0.56	0.0026
4	33.8	10	0.60	0.0036
5	32.0	9	0.39	0.0023
Average	33.2 ± 0.51	9.4 ± 0.40	0.54 ± 0.04	0.0031 ± 0.0003
Nano-TiO2 Exposure
1	34.7	10	0.61	0.0038
2	36.0	10	0.66	0.0043
3	35.6	9	0.66	0.0039
4	31.7	10	0.34	0.0026
5	35.7	9	0.61	0.0037
Average	33.7 ± 0.79	9.6 ± 0.24	0.58 ± 0.06	0.0037 ± 0.0003
P-Value	0.135	0.681	0.669	0.175

Engineered Nanomaterial (ENM) Inhalation Exposure

Our laboratory has previously detailed the engineered nanomaterial (ENM) inhalation exposure paradigm implemented in this study (Hathaway et al. 2019; Kunovac et al. 2021; Kunovac et al. 2019). While the concentration of the exposures are higher than the National Institute for Occupational Safety & Health (NIOSH) Recommended Exposure Limits (REL) for ultrafine TiO₂ (0.3 mg/m³), we are attempting to deliver a representative alveolar dose that is similar to what a working pregnant female would experience. Our dose is equivalent to a lightly working human female exposed to TiO₂ levels previously measured in factories (~0.65 mg/m³) for a period of approximately 200 days (40 weeks x 5 working days), which spans a full-term pregnancy.

Nano-TiO₂ P25 powder was purchased from Evonik (Aeroxide TiO₂, Parsipanny, NJ) which was composed of anatase (80%) and rutile (20%) TiO₂. Prior to aerosolization, the nano-TiO₂ powder was prepared by drying, sieving, and storing (Hathaway et al. 2019; Knuckles et al. 2012; Kunovac et al. 2021; Kunovac et al. 2019; Nurkiewicz et al. 2008). Previous studies from our laboratory and colleagues have outlined the primary particle characteristics including the size (21 nm), Zeta potential (−56.6 mV), and the specific surface area (48.08 m²/g) (Nichols et al. 2018; Nurkiewicz et al. 2008; Sager et al. 2008).

A high-pressure acoustical generator (HPAG; IEStechno, Morgantown, WV) was utilized for nano-TiO₂ aerosolization as has been previously done for rodent inhalation exposure studies (Hathaway et al. 2019; Kunovac et al. 2021; Kunovac et al. 2019). Figure 1 outlines the details of nano-TiO₂ aerosol characterization. Using a whole-body exposure chamber, a target aerosol mass concentration of 12 mg/m³ of engineered nano-TiO₂ was implemented for a period of 360 minutes per day for 6 non-consecutive days, over an 8-day period (Kunovac et al. 2021). The relevance of this concentration lies in its ability to recapitulate the lung burden of a person who works in a manufacturing setting, based on the human equivalent alveolar doses during pregnancy and has been previously detailed by our laboratory (Kunovac et al. 2021). Alveolar deposition fraction (F), minute ventilation based on body weight (V), the mass concentration (C), and exposure duration (T) were used with the equation D=F x V x C x T. This allowed us to determine that the daily deposited nano-TiO₂ alveolar dose was 6.92 μg (total six exposure dose = 41.55 μg).

Figure 1: Characteristics of maternal whole-body inhalation exposure.

(A) A target aerosol mass concentration of 12.01 ± 0.50 mg/m³ of engineered nano-TiO₂ was implemented for a period of 360 minutes per day for 6 non-consecutive days, over an 8-day period. These concentrations were verified by gravimetric measurements during each exposure, which resulted in an average mass concentration of 12.01 ± 0.50 mg/m³ during the 360-minute period. (B) High-resolution electrical low-pressure impactor (ELPI+) indicated a geometric count median diameter (CMD) of 172 nm with a geometric standard deviation (GSD) of 1.96. (C) Scanning particle mobility sizer (SMPS) indicated a CMD of 112 nm with a GSD of 2.14. (D) Exposure model paradigm including the allocation of tissue on gestational day (GD) 15.

The real-time TiO2 aerosol mass concentration readings were sampled from the exposure chamber during a typical exposure day (Figure 1A). These concentrations were verified by gravimetric measurements during each exposure, which resulted in an average mass concentration of 12.01 ± 0.50 mg/m³ during the 360-minute period. To determine the size distribution of the nano-TiO2 aerosols, a high-resolution electrical low-pressure impactor (ELPI+; Dekati, Tampere, Finland), and a scanning particle mobility sizer (SMPS 3938; TSI Inc., St. Paul, MN) were employed. The ELPI+ data indicated a geometric count median diameter (CMD) of 172 nm with a geometric standard deviation (GSD) of 1.96 (Figure 1B). The SMPS data indicated a CMD of 112 nm with a GSD of 2.14 (Figure 1C). The exposure chamber contained bedding material that was soaked in water to maintain a comfortable humidity level during exposure. Control animals (sham) were exposed to HEPA filtered air with similar chamber conditions in a designated control chamber. The final exposure was administered 48 hours prior to sacrifice and tissue harvesting.

RNA Isolation and Fragmentation

Total RNA was extracted from pooled fetal heart samples (sham, n =3; nano-TiO₂, n=3) using QIAzol lysis reagent (item no. 79306, Qiagen). Samples were homogenized in QIAzol using a rotor-stator homogenizer and further processed using the miRNeasy Mini Kit (item no. 217004, Qiagen) per manufacturer’s protocol. Concentrations were determined for each sample using the NanoDrop ND-100 (Thermo Fisher Scientific, Waltham, MA). Spike-in controls obtained from the EpiMark N6-Methyladenosine Enrichment Kit (item no. E1610S; NEB) were prepared following the manufacturer’s protocol. Briefly, the m⁶A control RNA contains m⁶A modified RNA (Gaussia luciferase), which was transcribed in the presence of 20% m⁶ATP and 80% ATP and the unmodified control RNA (Cypridina luciferase) contains no modifications. 1 μL of the m⁶A control RNA and 1 μL of the unmodified control RNA were added to each sample prior to fragmentation. The fetal cardiac total RNA samples (3 μg) were fragmented into ~250-350-nucleotide fragments using the NEBNext Magnesium RNA Fragmentation kit (item no. E6150S; New England Biolabs (NEB), Ipswich, MA) at 65°C for 5 minutes, followed by addition of the stop solution (Liu et al. 2020; Zeng et al. 2018), previously shown (Kunovac et al. 2021). The Monarch RNA Cleanup Kit (item no T2030L; NEB) was performed per manufacturer’s protocol. Samples were eluted with 20 μL of nuclease-dree water and concentrations were determined using the NanoDrop ND-100 (Thermo Fisher Scientific). 2 μL of fragmented total RNA was saved from each sample to be used as input and the rest was used for m⁶A RNA immunoprecipitation followed by sequencing (m⁶A-RIP-seq).

Immunoprecipitation (m⁶A-RIP) with Low Input Samples

The protocol used in our study was adapted from m⁶A-RIP protocols previously described (Zeng et al. 2018) (Liu et al. 2020), with modifications. In a 1.5 mL microcentrifuge tube (per sample), 30 μL of protein-A magnetic beads (item no. 10002D; Thermo Fisher Scientific) and 30 μL of protein-G magnetic beads (item no. 10004D; Thermo Fisher Scientific) were mixed, a magnetic field applied, and supernatant removed. The mixture of beads was washed twice with 400 μL of IPP buffer (10 mM Tris-HCl, 150 mM NaCl, pH 7.5; 0.1% IGEPAL CA-630) and resuspended in 500 μL of IPP buffer. 5 μg of affinity purified anti-m⁶A polyclonal antibody (item no. ABE572-I; MilliporeSigma, Burlington, MA) was added to the beads and incubated overnight at 4°C. An additional tube with the mixture of beads received 5 μg of the normal rabbit IgG control antibody (item no. 2729; Cell Signaling Technology (CST), Danvers, MA) and was also incubated overnight at 4°C. The magnetic field was applied, and supernatant removed, followed by washing of the antibody-bead mixture twice with IPP buffer. In a 1.5 mL microcentrifuge tube, a 500 μL mixture was prepared containing 100 μL of 5X IPP buffer, fragmented total RNA (diluted with nuclease-free water to a volume of 395 μL), and 5 μL of RNasin Plus RNase Inhibitor (item no. N2611; Promega, Madison, WI).

The antibody-bead mixture was resuspended in the 500 μL mixture containing the fragmented RNA and incubated with orbital rotation at 4°C for 2 hours, followed by application of the magnetic field, removal of the supernatant, and two washes with 1,000 μL of IPP buffer. The bead-antibody-RNA mixture was then washed twice with 1,000 μL of low-salt IPP buffer (10 mM Tris-HCl, 50 mM NaCl, pH 7.5; 0.1% IGEPAL CA-630) and twice with 1,000 μL of high-salt IPP buffer (10 mM Tris-HCl, 500 mM NaCl, pH 7.5; 0.1% IGEPAL CA-630). A competitive elution buffer was then prepared containing 6.7 mM N⁶-methyladenosine 5’- monophosphate sodium salt (item no. M2780, Sigma-Aldrich) in IPP buffer. Each immunoprecipitation (IP) sample was resuspended in 100 μL of the m⁶A competitive elution buffer with continuous shaking at 4°C for 1 hour. The magnetic field was applied, which allowed us to collect the supernatant (eluted m⁶A RNA) into a new tube. The eluted RNA was further purified using the RNeasy MinElute Cleanup Kit (item no. 74204; Qiagen) following manufacturer’s protocol. The m⁶A enriched RNA was eluted by adding 14 μL of ultrapure H₂O directly to the center of the membrane and centrifuging for 1.5 mins at full speed. Samples were stored at −80°C until initiation of the library preparation procedure.

Library Prep and Sequencing

Library preparation was performed on immunoprecipitated (IP) and total (input) RNA samples from sham (n = 3) and nano-TiO₂ (n = 3) gestationally-exposed fetal offspring hearts using the NEB^® Single Cell/Low Input RNA Library Prep kit for Illumina^® (item no. E6420S, NEB) and Illumina^® compatible NEB^® UDIs (item no. E6440S, NEB), which mitigate sample misassignment due to index-hopped reads. The Agilent 2100 BioAnalyzer (Agilent Technologies, Santa Clara, CA) was employed to determine the size distribution of RNA. Libraries were PCR amplified for 10 cycles and NEB UDIs were used for indexing by amplifying for 8 cycles. The RNA sample concentrations were quantified using a Qubit Fluorometer (Thermo Fisher Scientific). The samples were sequenced using the Illumina NextSeq 2000 (Illumina Inc., San Diego, CA) as paired end (PE) 2x50 bp reads.

Analysis of RNA-Seq Data

Adapters were trimmed from Fastq files through Flexbar v3.5 (Roehr et al. 2017). The trimmed Fastq files were processed using paired-end alignment with HISAT2 v2.2.1 and the resulting BAM files were aligned to the mouse reference genome (Mus musculus, GRCm39, Ensemble version 104) for input and IP samples. Peaks showing significant enrichment in the IP samples vs. corresponding input samples for all submitted replicates were detected using RADAR v0.2.1 (Zhang et al. 2019), with a fragment length of 297 base pairs and bin size of 25 base pairs. Peak reads coverage for transcripts were visualized with RADAR and the Integrative Genomics Viewer (IGV) browser (Berulava et al. 2019; Engel et al. 2018; Liu et al. 2019; Robinson et al. 2011).

Input mRNA samples were further processed for full decoy-aware transcriptomic analyses using Salmon v1.5.2 (Patro et al. 2017) on the mouse reference genome (Mus musculus, GRCm39, Ensemble version 104) and processed in R using tximport v1.22.0 (Soneson et al. 2015). Differential gene expression was performed through DESeq2 v1.34.0 (Love et al. 2014). Visualization of data was accomplished through ggpolt2 v3.3.5 (Wickham 2016) and pheatmap v1.0.12. Pathway analysis was performed through PathFindR v1.6.3 (Ulgen et al. 2019).

Mitochondrial Isolation

Fetal pups were humanely killed at GD 15 and their hearts were excised through a midsagittal cut in the thoracic cavity. Mitochondria were isolated as previously described (Palmer et al. 1977), with modifications by our laboratory (Baseler et al. 2013; Baseler et al. 2011; Dabkowski et al. 2010). Briefly, whole hearts were gently homogenized using a dounce homogenizer by hand. Homogenates were centrifuged at 700 g for 10 min. The supernatant was extracted (containing the subsarcolemmal mitochondria) and centrifuged again at 10,000 g to isolate the subsarcolemmal mitochondrial fraction. Additional washes and centrifugation steps at 10,000 g were implemented to purify the subsarcolemmal mitochondrial fraction. Following the first 700 g centrifugation, the interfibrillar (remaining after the subsarcolemmal mitochondria supernatant was removed) was treated with trypsin (5 mg/g) for 10 minutes. Differential centrifugation steps at 10,000 g were further applied to isolate and purify the interfibrillar mitochondrial fraction. The two subpopulations of mitochondria (subsarcolemmal and interfibrillar) were combined at the end of the protocol to achieve a total mitochondrial population. Samples were resuspended in KME buffer (100mM KCl, 50mM MOPS and 0.5mM EGTA pH 7.4) with protease inhibitor cocktail (PIC).

Sample Preparation for Mitochondrial Proteomics

Label-free proteomics sample preparation was performed on fetal cardiac isolate mitochondria (sham, n=4; nano-TiO₂, n=4) as described with several modifications (Valentine et al. 2009). 37 μL, containing about 24 μg of protein, as measured by the Bradford assay, of isolated fetal cardiac mitochondria (in KME+PIC) was placed in a SpeedVac for 3.5 hours to dry out the sample. Samples were resuspended in 20 μL of Dissolution buffer (8 M Urea, 50 mM Tris-HCl - pH 8.2)) and their protein concentrations were determined using the NanoDrop^™ One Microvolume UV-Vis Spectrophotometer (Thermo Fisher Scientific). 12 μg of protein per sample was used for the rest of the sample preparation procedure. 1.2 μL of DTT reducing agent (100 mM DTT (item no. V3151, Promega) in 8 M urea and 100 mM ammonium bicarbonate – pH 8.2) was added to each sample followed by a 2-hour incubation at 37°C, with gentle agitation. After reducing the disulfide bonds, samples were placed on ice. Protein alkylation was then achieved using 1.2 μL of iodoacetamide (225 mM iodoacetamide (item no. I6125, Sigma-Aldrich) in 8 M urea and 100 mM ammonium bicarbonate – pH 8.2), which was made fresh.

The reaction was allowed to proceed in darkness, on ice for 2 hours. Next, 1.2 μL of cysteine blocking reagent (100 mM L-cysteine (item no. 168149, Sigma-Aldrich) in 8 M urea and 100 mM ammonium bicarbonate – pH 8.2) was added to the mixture to remove any remaining reagent, followed by a 30-minute incubation at room temperature with gentle agitation. Following this reaction, enough Dilution solution (100 mM ammonium bicarbonate – pH 8.2) was added to each sample to bring the urea to a 2 M concentration. Protein digestion was initiated by the addition of Trypsin solution (appropriate amount of trypsin in 2 M urea, 100 mM ammonium bicarbonate – pH 8.2) to a final concentration of 50:1 protein:trypsin and incubated for 24 hours at 37°C, with gentle agitation. Peptide cleanup was performed using Thermo Scientific Pierce^™ C18 Spin Columns (item no. 89873) per manufacturer’s instructions and samples were eluted using acetonitrile. Samples were dried using a centrifugal concentrator (Labconco) and stored at −20°C for LC-MS/MS analysis. Prior to LC-MS/MS analysis, each sample was resuspended in 100 μL of formic acid buffer. Due to limited quantities of protein, the four samples in each group were pooled to produce sham (n=2) and nano-TiO₂ (n=2).

LC Separation

20 μL of digest solution was injected onto a reversed phase liquid chromatography (column (10 cm × 2.1 mm ID, 5 μm particle dia). A flow rate of 300 μL/min was employed for the reversed-phase separation. Gradient elution was performed using two buffers. Buffer A consisted of HPLC-grade water with 1% formic acid and Buffer B was acetonitrile with 1% formic acid. During LC separations, the solvent contribution from Buffer B was 0.1%, 0.1%, 25%, 80%, 90%, 90%, and 0.1% at 0, 2, 27, 37, 42, 52, and 57 minutes, respectively.

Mass Spectrometer

Data were collected on an orbitrap mass spectrometer (Q-Exactive, Thermo Scientific). Experiments were conducted in positive ion mode and data dependent analysis was employed. Throughout the LC separation, a full MS scan was collected followed by 5 MS2 scans. Resolving power settings of 70,000 and 17,500 were utilized for the precursor and MS2 scans, respectively. The respective AGC settings were 1E6 and 1E5. A normalized collision energy of 30 was employed for MS2 analyses. Finally, a dynamic exclusion time of 10 s was employed for the separations.

Data Processing

Raw data files from the Q-Exactive mass spectrometer were imported into the Proteome Discoverer software suite (Thermo Scientific). The SEQUEST search engine was employed for peptide identification. The SwissProt (v2017-10-25) database for Mus musculus was employed for peptide ion identification. The search employed a minimum peptide length of 5 residues and allowed for 3 missed cleavages using Trypsin as the enzyme used for digestion. A precursor mass tolerance of 12 ppm and a fragment ion mass tolerance of 0.03 Da was employed. A static modification of 57.021 Da (C, carbamidomethyl)) was employed in the searches. A value of 0.1 was used for protein and peptide validation settings.

Proteomic Analysis

After the results were obtained, analysis was performed using an exponentially modified protein abundance index $(\text{emPAI})$ to estimate absolute protein amounts based on the number of sequenced peptides per protein (Ishihama et al. 2005). Briefly, only high-quality peptides were selected based on the Score Sequest HT: A Sequest HT and mitochondrial proteins were identified using MitoCarta3.0 (Rath et al. 2021). The protein abundance index $(\text{PAI})$ was calculated with Equation 1 using the “total peptides” count per protein for $N_{\text{observed}}$ and the $N_{\text{observable}}$ was determined using ExPASy server (Wilkins et al. 1999). The emPAI was then calculated as shown in Equation 2. The $\text{emPAI}$ values for the two samples per group was averaged for each respective group (sham and nano-TiO₂) and compared.

$$PAI=\frac{N_{\text{observed}}}{N_{\text{observable}}}$$ Equation 1

$$emPAI={10}^{\text{PAI}}-1$$ Equation 2

Statistics

Statistical analyses were conducted in the R (v4.1.1) environment and GraphPad Prism (v9.5.1). A 2-sided Student’s t-test was used for two normally distributed, continuous variables. The Shapiro-Wilk test was utilized for normality testing. For non-Gaussian distributed data, the Mann-Whitney U test was applied. DESeq2 implements the Wald Test, using the Benjamini-Hochberg procedure. The FDR for this study was set to 0.05 for the transcriptomic analyses and 0.10 for m⁶A peak calling. All measures of significance are reported as adjusted P-values (Padj). The RADAR package used for peak calling implements a Gamma-Poisson Distribution Model, which is a type of random effect model.

Results

Study Design

The real-time TiO₂ aerosol mass concentration readings were sampled from the exposure chamber during a typical exposure day demonstrating an average mass concentration of 12.01 ± 0.50 mg/m³ (Figure 1A) and geometric count median diameter of 172 nm (Figure 1B, ELPI+) and 112 nm (Figure 1C, SMPS). The study design (Figure 1D) highlights the use of mRNA and m⁶A sequencing on whole heart tissue, with proteomics specifically on isolated mitochondria. Maternal dams were humanely killed at gestational day (GD) 15. Each dam was considered an n=1. No significant differences were observed between maternal and fetal pup weights, pup number, or pup heart weights (Table 1).

Transcriptomics

In performing the m⁶A-RIP-Seq, the input controls provided overall transcript abundance, which showcased the changes to mRNA seen following gestational nano-TiO₂ exposure in fetal offspring. Sample distances were projected both using multidimensional scaling (MDS) (Figure 2A) as well as distribution with variance stabilizing transformation (VST) (Figure S1A). Both techniques illustrate distinct stratification between the two groups. An illustration of the VST, as well as Log2 normalized counts and regularized-logarithm (rlog) transformations, are provided (Figure S1B). Differential expression analysis was performed and revealed a total of 3,648 transcripts that were significantly changed following exposure (Figure 2B). Examining the top 1,000 transcripts, sorted by Padj value, we see that 2/3 are upregulated, while 1/3 are downregulated in the nano-TiO₂ group compared to sham (Figure 2C). Additionally, Figure 2C shows that the sham and nano-TiO₂ groups hierarchically cluster separately.

Figure 2: Transcriptomic analysis of fetal cardiac tissue.

(A) Multidimensional Scaling (MDS) plot of sham (n=3) and nano-TiO₂ (n=3) gestationally exposed fetal progeny. (B) Volcano plot depicting the Log2 fold change observed following differential expression analysis. The top gene (lowest Padj value) is listed. (C) Heatmap depicting the top 1,000 genes, as sorted in ascending order by Padj value. The provided scale is in Log2 fold change. (D) Heatmap of differentially expressed mitochondrial genes within the electron transport chain oxidative phosphorylation pathway. Genes were classified based on MitoCarta3.0 (Rath et al. 2021). The provided scale is in fold change. Genes were differentially expressed if Padj≤0.05, FDR=0.05. Differences are illustrated as nano-TiO₂ (n=3) compared to sham (n=3). Sham = fetal progeny of maternal dams exposed to filtered air, Nano-TiO₂ = fetal progeny of maternal dams exposed to 12.01 ± 0.50 mg/m³ of nano-TiO₂ for 360-minute periods for 6 days, Complex I = NADH:ubiquinone oxidoreductase, Complex II = succinate dehydrogenase, Complex III = coenzyme Q – cytochrome c reductase, Complex IV = cytochrome c oxidase, ATP Synthase = mitochondrial electron transport chain complex V.

Our previous investigations have shown that mitochondrial dysfunction, specifically alterations to bioenergetics, are correlated with cardiovascular functional impairments in fetal and neonatal offspring, such as reduced ejection fraction, changes to left ventricular size, and alterations of strain of the heart (Hathaway et al. 2017; Kunovac et al. 2021; Kunovac et al. 2019). Transcription of electron transport chain proteins were significantly decreased for all mitochondrial complexes in the nano-TiO₂ group (Figure 2D). Decreased expression was seen across most categories of mitochondrial genes (Figure S2). Interestingly, mitochondrial electron transport chain complex genes (Figure 2D) and mitochondrial ribosomal metabolism (Figure S2) were the two pathways most significantly impacted following nano-TiO₂ exposure. Also of interest, mitochondrial glutathione peroxidase 4 (GPx4) (Figure S1C) was significantly decreased, which we have reported previously. Our previous studies have indicated that GPx4 is an important mediator of ROS levels in nano-TiO₂ gestationally and directly exposed mice (Kunovac et al. 2021; Nichols et al. 2018).

Epitranscriptomics through N⁶-methyladenosine (m⁶A)

Using both the input (unmodified mRNA) and immunoprecipitated RNA with our m⁶A antibodies, we wanted to examine if specific RNA sites were being methylated, or demethylated, following nano-TiO₂ exposure (Figure 3). Principal Component Analysis (PCA) revealed a distinct separation between sham and nano-TiO₂ groups (Figure 4A). Through differential peak analyses, we determined that the majority of m⁶A sites were occurring in 3’-untranslated regions (UTRs) (Figure 4B), which is supported by previous literature (Meyer et al. 2012). We then plotted all the differentially methylated sites, revealing significant upregulation of m⁶A methylation following exposure (Figure 4C).

Figure 3: Schematic overview of N⁶-methyladenosine (m⁶A) analyses.

First, RNA was isolated from fetal pup hearts. Next, RNA was fragmented, with an average fragment size of 297 base pairs. 10% of each sample was saved as an input control while 90% of the sample was immunoprecipitated with m⁶A antibodies. RNA was barcoded and reverse transcribed into cDNA. Sequencing was performed on the Illumina NextSeq with paired-end (PE) 2x50 reads. Adapters were trimmed from Fastq files, files were aligned to the mouse genome, and differential peak calling was performed. Sham = fetal progeny of maternal dams exposed to filtered air, Nano-TiO₂ = fetal progeny of maternal dams exposed to 12.01 ± 0.50 mg/m³ of nano-TiO₂ for 360-minute periods for 6 days, Input = unprocessed mRNA, IP = mRNA derived from immunoprecipitation with m⁶A antibodies.

Figure 4: Epitranscriptomic analysis of fetal cardiac tissue.

(A) Principal Component Analysis (PCA) plot of sham (n=3) and nano-TiO₂ (n=3) gestationally exposed fetal progeny. (B) Pie chart depicting the distribution of m⁶A sites identified by location across genomic space. (C) Heatmap of the 1,135 differentially m⁶A methylated sites across 921 genes. Z-Score is defined as (gene expression value in sample of interest) – (mean expression across all samples)/standard deviation. Bins were defined as 25 base-pair regions. (D) Heatmap of the shared 331 genes that were both identified in the transcriptomic and epitranscriptomic analyses, with their respective mRNA and m6A expression levels. (E) Heatmap of mitochondrial oxidative phosphorylation genes with their respective mRNA and m6A expression levels. The provided scale for (D) and (E) is in fold change. Genes were differentially expressed if Padj≤0.10, FDR=0.10. Sham = fetal progeny of maternal dams exposed to filtered air, Nano-TiO₂ = fetal progeny of maternal dams exposed to 12.01 ± 0.50 mg/m³ of nano-TiO₂ for 360-minute periods for 6 days, 3’ UTR = 3’ untranslated region of a gene, 5’ UTR = 5’ untranslated region of a gene, CDS = coding DNA sequence, ncRNA = non-coding RNA, Other = other regions in genomic space not specified.

Our next step was to assess if the mRNA identified in the input samples and the differentially methylated sites overlapped. Of the 3,648 differentially expressed mRNA transcripts and 921 unique transcripts containing significant m⁶A modifications, 311 overlapped (Figure 4D). In these 311 genes, mRNA expression was primarily increased (277 of 311 genes), suggesting a predominately activating effect of m⁶A on gene expression/transcript longevity. Likewise, the mitochondrial genes, revealing little to no change in m⁶A expression, had an overall reduced mRNA expression (Figure 4E).

mRNA/m⁶A Shared Features

We next wanted to understand how the 311 genes identified in both mRNA and m⁶A analyses were related to cellular and molecular pathways. Using gene ontology through Kyoto Encyclopedia of Genes and Genomes (KEGG), we hierarchically clustered the pathways and represented the top genes within the pathways for the 3,648 differentially expressed mRNA transcripts (Figure 5A) as well as the 311 shared genes (Figure 5B). Comparing all transcripts in Figure 5A to the shared genes in Figure 5B there were common cellular and molecular pathways that persisted, including ubiquitin-mediated proteolysis, RNA degradation, spliceosome, and hypoxia inducible factor (HIF)-1 signaling. Interestingly, of the 3,648 differentially expressed mRNA transcripts (Figure 5A), oxidative phosphorylation was one of the top pathways, though transcripts involved in oxidative phosphorylation did not have significant m⁶A modifications following exposure (Figure 5B). The common pathways between the two groups were mainly larger regulator pathways, controlling the transcription, translation, and turnover of proteins. We include the detailed molecular pathway for RNA degradation of all 3,648 differentially expressed transcripts (Figure S3A) as well as for the shared mRNA and m⁶A differentially expressed genes (Figure S3B).

Figure 5: Pathway analysis of differentially expressed genes.

(A) Pathway analysis of the 3,648 genes differentially expressed in the transcriptomic analysis. (B) Pathway analysis of those genes (311) that also coincided with differentially methylated sites. Both pathways are presented as hierarchically clustered pathways, with the top (A) six and (B) four pathways represented along with other similar pathways within each clustered group. Genes were differentially expressed if Padj≤0.05, FDR=0.05

Mitochondrial Proteomics

Our last objective was to analyze the mitochondrial proteome to define if transcriptional alterations were associated with protein level changes. Our workflow included isolating mitochondria from fetal heart tissue, isolating the protein, and processing the peptides through LC-MS/MS (Figure 6A). The exponentially modified protein abundance index (emPAI) was calculated for mitochondrial proteins identified with a high confidence in both the sham and nano-TiO₂ groups. Of the mitochondrial proteins evaluated, those belonging to ATP Synthase, including Atp5a1, Atp5b, Atp5c1, Atp5pb, and complex I, Nudufs1 and Nudufv2, showed the most consistent downregulation in the nano-TiO₂ exposure group (Figure 6B). Additionally, peroxiredoxin 3 (Prdx3), a critical mitochondrial antioxidant, revealed the largest decrease between the sham and exposed groups (Figure 6B). Predominately, proteins involved in mitochondrial bioenergetics were decreased following gestational nano-TiO₂ exposure.

Figure 6: Mitochondrial proteomics.

(A) Schematic depicting the processes of mitochondrial isolation, protein extraction, peptide preparation, and peak identification. (B) Mitochondrial proteins that were identified with high confidence in the analyses. The provided scale is the difference between emPAI values between the sham and nano-TiO₂ groups. Proteins are organized based on MitoCarta3.0 designations (Rath et al. 2021). Sham = fetal progeny of maternal dams exposed to filtered air, Exp = fetal progeny of maternal dams exposed to 12.01 ±0.50 mg/m³ of nano-TiO₂ for 360-minute periods for 6 days, LC-MS/MS = liquid chromatography with tandem mass spectrometry, emPAI = exponentially modified protein abundance index.

Discussion

Fetal development is an intricate process that requires a significant number of regulatory mechanisms. A baleful gestational environment can result in disease predisposition in addition to immediate consequences of fetal development. M⁶A can transiently alter cellular function and may play a significant role in the host response to toxicants during gestation. The results of our study emphasize the impact of m⁶A methylation on regulatory pathways within the heart, largely involved with the transcription, translation, and longevity of transcripts and proteins within the cell and how genes with no change in m⁶A methylation (i.e., nuclear-transcribed mitochondrial genes) have decreased expression following exposure. Along with changes to oxidative phosphorylation, acute changes to the m⁶A methylome may predispose offspring to future adverse outcomes. Ultimately, maternal inhalation exposure to nano-TiO₂ significantly alters both transcription and translation of key genes within the fetus.

Exposure to environmental toxicants is associated with developmental reprogramming that alters normal physiological responses and thereby leads to disease susceptibility including metabolic diseases (Trevino et al. 2020; Walker 2011). Specifically, prenatal maternal exposure to benzene, diesel exhaust, and fine particulate matter lead to glucose intolerance and elevated insulin resistance (Koshko et al. 2021), weight gain (Bolton et al. 2012), and metabolic syndrome (Wu et al. 2019), respectively, in adulthood. As cardiovascular disease remains the leading cause of death worldwide, researchers have begun investigating whether m⁶A mRNA modifications can serve as diagnostic or therapeutic targets (Komal et al. 2021; Sweaad et al. 2021). Transcriptome-wide assessments of cardiac samples from human and murine models of dilated cardiomyopathy were assessed by Kmietczyk et. al (Kmietczyk et al. 2019). They found that hypermethylation was prominent in failing human hearts and confirmed the dynamic and regulatory capabilities of m⁶A in this diseased state.

Understanding how the results of our study (i.e., increased m⁶A methylation following nano-TiO₂ exposure) can be generalized to broader inhaled environmental toxicants is important. Similar to nano-TiO₂, global m⁶A hypermethylation in peripheral blood mononuclear cells (PBMCs) was found in students with known PM2.5 exposure (Li et al. 2023). A positive association was also seen in the Beijing Truck Driver Air Pollution Study between carbon black and increased global m⁶A in blood (Kupsco et al. 2020). While not m⁶A specific, DNA hypermethylation of insulin-like growth factor 2 mRNA-binding protein 1 (IGF2BP1) was associated with recurrent spontaneous abortion in decidual tissue from women exposed to a variety of environmental toxins, including PM₁₀, PM_2.5, CO, NO₂, and SO₂ (Zhu et al. 2022). These studies suggest increased methylation, whether deleterious or compensatory, is a major contributor to modifications of genes following exposure; direct changes to gene transcription, specifically in the heart following carbon black intratracheal instillation, does not appear to have a significant contribution (Bourdon et al. 2013). The mechanisms driving the upregulation of epigenetic machinery is likely due to indirect inflammatory processes involving cytokines and toll-like receptors in the placenta, which can mediate cellular reprograming and ROS generation (Hougaard et al. 2015).

The mechanistic role of ROS-mediated, global m⁶A activation is probably best captured by Ruan et al. (Ruan et al. 2022). Following in vitro (i.e., A549 cells and Beas-2B cells) and in vivo (i.e., participants blood samples) nano-TiO₂ exposure, the authors revealed methyltransferase-like 3 (METTL3) stabilization and increased expression (i.e., more methylation activity) resulting in ERK1/2 activation, m⁶A upregulation, and an inflammatory response that was ultimately driven by ROS (Ruan et al. 2022). While increased global mRNA m⁶A methylation has been demonstrated in multiple studies (Kupsco et al. 2020; Li et al. 2023; Ruan et al. 2022; Zhu et al. 2022), some studies have suggested a global hypomethylation (Cayir et al. 2019). Following exposure to PM2.5, there was a global decrease in m⁶A methylation, concomitant with elevated expression of methyltransferase complex components (METTL3 and Wilm’s Tumor 1-associating protein (WTAP)), demethylases (FTO and ALKBH5), and the reader protein heterogeneous nuclear ribonucleoproteins C1/C2 (HNRNPC) (Cayir et al. 2019). Additionally, the role of m⁶A methylation on other RNA species (such as non-coding RNAs) has been relatively unexplored. Han et al. demonstrated that the PI3K/AKT/mTOR pathway is altered in rat lung with inhaled carbon black exposure (Han et al. 2020). Following exposure, m⁶A of pri-miRNA-126 was decreased, correlating with increased pri-miRNA-126 accumulation, decreased mature miRNA-126, activation of PI3K/AKT/mTOR pathway, and ultimately increased fibrosis in the lung (Han et al. 2020).

Recent work from our laboratory has demonstrated that maternal inhalation exposure to ENM during a critical point in fetal development results in reduced cardiac contractility, mitochondrial bioenergetic disruption, and an increase in global m⁶A content (Kunovac et al. 2021). Notably, m⁶A methylation levels were reduced in the presence of enhanced antioxidant capacity that was mitochondrially targeted in the pregnant mother. It has become clear that a specific and balanced m⁶A RNA distribution is crucial to maintain homeostasis and evade disease, though the precise mechanisms underpinning the process remain unknown. Zhang et. al. similarly investigated the effects of carbon black nanoparticle exposure during pregnancy on m⁶A status in offspring, but exclusively from the perspective of neurobehavior deficits (Zhang et al. 2020). The authors determined that the exposure decreased m⁶A modification in the cortex, along with neurobehavioral deficits and cortex injuries in the offspring. These studies substantiate the idea that m⁶A methylation status is likely disease and tissue specific.

Following gestational exposure, mitochondria can develop longstanding, detrimental adaptations that place fetal offspring at risk. During early development, programming, and reprogramming, of mitochondria can be attributed to changes in mitochondrial copy number, oxidative phosphorylation efficiency, electron transport chain complex activities, and production of ROS that can carry over into adulthood (Gyllenhammer et al. 2020). While mitochondria have degrees of plasticity, the more damage incurred from external sources (e.g., ROS, inflammatory medicators, direct translocation of ENM), the more susceptible the cell is to developing DNA damage and epigenetic modifications to nuclear-encoded mitochondrial genes (Gyllenhammer et al. 2020) and promoting the integrated stress response (Mick et al. 2020). Mitochondrial reprograming in fetal progeny may not become evident until additional insults occur later in life (Gyllenhammer et al. 2020), leading to insufficient adaptive responses to the insult, which may even carry sex-dependent consequences (Griffith et al. 2023). We have previously shown increases in ROS in fetal and adult progeny following gestational ENM exposure (Kunovac et al. 2021; Kunovac et al. 2019), correlating with epigenetic and epitranscriptomic changes. The current study further elucidates how maladaptive m⁶A modifications can be linked to decreased mitochondrial protein translation in fetal progeny following exposure, contributing to known mitochondrial bioenergetic deficits (Kunovac et al. 2019).

The goal of our study was to profile the fetal cardiac m⁶A methylome and understand its contribution to known disruptions in mitochondrial bioenergetics. While we highlight m⁶A methylation is predominately occurring in pathways altering transcription of mRNA (spliceosome pathway), stability of transcripts (RNA degradation pathway), and protein longevity (ubiquitin-mediated proteolysis pathway), this does not directly define a mechanistic pathway to alter nuclear-transcribed mitochondrial gene expression. Instead, we propose that the m⁶A levels seen on nuclear-transcribed mitochondrial mRNA transcripts is the result of a maladaptive response to ENM exposure. M⁶A modifications in the 3’ untranslated region of mRNA can act to stabilize the transcript (Edupuganti et al. 2017; Huang et al. 2018), as observed in this study. By increasing expression of pathways such as RNA degradation, this would place mRNA transcripts at risk of overall shorter half-lives. To overcome the increase in mRNA turnover this would necessitate a concomitant increase in m⁶A, to stabilize the mRNA transcript and avoid premature degradation. The decreased expression of mitochondrial transcripts and protein in the current study could be a direct effect of this interplay, as mitochondrial mRNA showed no change in m⁶A expression, therefore allowing mRNA degradation to be the prominent pathway.

A limitation to the current study is the need to pool fetal hearts for analyses. To obtain a sufficient quantity of mitochondria, fetal hearts were pooled. Kit-based mitochondrial isolations recommend 0.3-0.5 grams of starting tissue, while tissue dissociation and differential filtration/centrifugation methods (such as the methodology used in the current study) recommend starting concentrations of ~0.2 grams wet weight (Preble et al. 2014). For this reason, sex-based differences were not characterized in the current study, as the tissue quantity was very limited. In the future, examining either later timepoints (> GD 15) or individually sexing fetal mice may help to provide sufficient tissue for sex-based analyses (Murdaugh et al. 2018). Another limitation is the single timepoint at GD 15. Our reasoning for choosing GD 15 was that GD 15 marks the end of cardiac organogenesis, which is the first time point in which all the cardiac structures have formed/matured; this includes separation of the aorta and the pulmonary artery (GD 13) and ventricular and atrial septation (GD 14.5) (Keller et al. 1996; Vuillemin and Pexieder 1989). An earlier timepoint would also make tissue collection difficult and, due to incomplete organogenesis, harvesting of cardiac tissue would be incomplete. Conversely, if a later timepoint was chosen, the epitranscriptomic mechanisms that are likely transient, may not be as readily observed.

Our current study, for the first time, explores the cardiac m⁶A methylome of fetal progeny following gestational inhalation exposure and its contribution in altering nuclear-transcribed mitochondrial transcripts. Implementing a multi-omics approach using transcriptomics, epitranscriptomics, and mitochondrial proteomics we detailed the shared pathways observed following nano-TiO₂ exposure. This framework helped to identify how m⁶A methylation can be globally increased in the heart following exposure and how lack of m⁶A expression contributes to decreased mitochondrial mRNA stability and, ultimately, decreased translation of mitochondrial proteins involved in bioenergetics. These data substantiate the need to further investigate RNA modifications as a mechanism for adaptive fetal responses and how these regulatory networks control nuclear-transcribed mitochondrial transcripts. The toxicity induced by gestational ENM exposure exerts multiple levels of transcriptional and translational regulation within the fetus and further highlights the direct, and indirect, impact of inhaled toxicants.