*Article* **Methylmercury-Induced Metabolic Alterations in** *Caenorhabditis elegans* **Are Diet-Dependent**

**Nicole Crawford 1, Megan Martell 1, Tyson Nielsen 1, Belal Khalil 1, Farooq Imtiaz 1, Etienne Nguidjo 1, Jennifer L. Newell-Caito <sup>2</sup> , Julia Bornhorst <sup>3</sup> , Tanja Schwerdtle <sup>4</sup> and Samuel W. Caito 1,\***


**Abstract:** Methylmercury (MeHg) is a well-known neurotoxicant; however, its role in metabolic diseases has been gaining wider attention. Chronic exposure to MeHg in human populations shows an association with diabetes mellitus and metabolic syndrome (MS). As the incidences of both obesity and MS are on the rise globally, it is important to understand the potential role of MeHg in the development of the disease. There is a dearth of information on dietary interactions between MeHg and lipids, which play an important role in developing MS. We have previously shown that MeHg increases food seeking behaviors, lipid levels, fat storage, and pro-adipogenic gene expression in *C. elegans* fed the standard OP50 *Escherichia coli* diet. However, we hypothesized that these metabolic changes could be prevented if the worms were fed a bacterial diet lower in lipid content. We tested whether *C. elegans* developed metabolic alterations in response to MeHg if they were fed two alternative *E. coli* strains (HT115 and HB101) that are known absorb significantly less lipids from their media. Additionally, to explore the effect of a high-lipid and high-cholesterol diet on MeHg-induced metabolic dysfunction, we supplemented the OP50 strain with twice the standard concentration of cholesterol in the nematode growth media. Wild-type worms fed either the HB101 or HT115 diet were more resistant to MeHg than the worms fed the OP50 diet, showing a significant right-hand shift in the dose–response survival curve. Worms fed the OP50 diet supplemented with cholesterol were more sensitive to MeHg, showing a significant left-hand shift in the dose–response survival curve. Changes in sensitivity to MeHg by differential diet were not due to altered MeHg intake in the worms as measured by inductively coupled mass spectrometry. Worms fed the low-fat diets showed protection from MeHg-induced metabolic changes, including decreased food consumption, lower triglyceride content, and lower fat storage than the worms fed either of the higher-fat diets. Oxidative stress is a common characteristic of both MeHg exposure and high-fat diets. Worms fed either OP50 or OP50 supplemented with cholesterol and treated with MeHg had significantly higher levels of reactive oxygen species, carbonylated proteins, and loss of glutathione than the worms fed the HT115 or HB101 low-lipid diets. Taken together, our data suggest a synergistic effect of MeHg and dietary lipid levels on MeHg toxicity and fat metabolism in *C. elegans*, which may affect the ability of MeHg to cause metabolic dysfunction.

**Keywords:** methylmercury; diet; cholesterol; high fat; low fat

**Citation:** Crawford, N.; Martell, M.; Nielsen, T.; Khalil, B.; Imtiaz, F.; Nguidjo, E.; Newell-Caito, J.L.; Bornhorst, J.; Schwerdtle, T.; Caito, S.W. Methylmercury-Induced Metabolic Alterations in *Caenorhabditis elegans* Are Diet-Dependent. *Toxics* **2021**, *9*, 287. https://doi.org/10.3390/toxics9110287

Academic Editors: Richard Ortega and Asuncion Carmona

Received: 1 September 2021 Accepted: 29 October 2021 Published: 2 November 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

### **1. Introduction**

Metabolic syndrome (MS) and obesity are major health concerns with increasing prevalence worldwide. MS is defined as a multifactorial condition characterized by insulin resistance, diabetes mellitus (DM), dyslipidemia, and obesity. It has become increasingly evident that many factors influence the prevalence of MS, including environmental factors [1,2]. One such environmental agent emerging as a potential obesogen is methylmercury (MeHg). MeHg is a well-known neurotoxin, which, in developmental exposures, causes cognitive and behavioral dysfunction in children and is linked to the development of neurodegenerative diseases, such as Parkinson's disease [3,4]. While currently being debated, there is growing evidence for a link between MeHg exposure and the development of MS. The National Health and Nutrition Examination Survey (NHANES) and Korean NHANES (KNHANES) data from 2003–2014 and 2011–2013, respectively, support an association between blood heavy metal levels (which include Hg) with MS, obesity, and lipid dysregulation [5–7]. These studies highlight the effect of heavy metals on the development of MS; however, it is unclear as to whether the observed metabolic effects were due to a single metal or a synergism of multiple metals in the mixture. However it has been shown that elevated blood mercury levels are also associated with increased visceral adipose tissue [8], and that toenail mercury levels (a marker for chronic Hg exposure) are associated with the development of MS [9]. We have recently shown that MeHg significantly increased lipid storage, altered feeding behavior, and increased the transcription of MS-related genes in *Caenorhabditis elegans* (*C. elegans*) [10]. Mechanisms that lead to MeHg-induced dyslipidemia are not known.

As a toxicant, MeHg primarily enters the human body through our diet. MeHg is a major contaminant of our fish supply, with greater Hg concentrations present in fish higher up the food chain, such as tuna, shark, swordfish, and mackerel [11]. While there are many dietary benefits from regular fish consumption, such as increased polyunsaturated fatty acids (PUFA) and selenium intake, the level of Hg ingested is an important consideration, especially for children and pregnant women. Studies have shown that dietary factors can affect how much Hg enters the body and its toxicity. MeHg enters cells through a molecular mimicry mechanism. MeHg readily binds to thiol groups. When bound to the amino acid cysteine, the MeHg–cysteine molecule resembles the amino acid methionine, and is able to enter cells through the large amino acid transporter 1 and 2 (LAT1 and LAT2) [12]. In worms, the amino acid transporters 1–3 are homologs to LAT1 and LAT2, which transport MeHg into cells [13]. If worms are fed a diet enriched in methionine, MeHg transport into cells is significantly decreased, as well as its toxic effects [13]. Similar effects have been observed in mammalian systems [14,15]. In addition to binding thiols, MeHg has high affinity for selenium. In diets enriched in selenium, MeHg will bind selenium and selenoproteins rather than thiols, preventing glutathione depletion and MeHg toxicity [16,17].

The relationship between dietary fats and methylmercury exposure has gained much attention. Polyunsaturated n-3 fatty acids, such as eicosapentaenoic acid (20:5n-3, EPA) and docosahexaenoic acid (22:6n-3, DHA), have multiple health benefits, from lowering serum low-density lipoprotein levels to being cardio-protective and preventing metabolic diseases [18,19]. These fatty acids are high in fish species that also contain significant Hg levels. Therefore, understanding the relationship between PUFA and Hg consumption is important. Longitudinal studies performed in the Seychelle Islands have shown beneficial effects on cognition in children from maternal exposure to PUFAs found in fish contaminated with MeHg [20–22]. n-3 PUFAS may protect against MeHg toxicity, either by decreasing apoptosis or by reducing MeHg uptake [23]. Interestingly, a meta-analysis has shown that high circulating n-3 PUFA levels or fish consumption correlate with a lower risk of developing metabolic syndrome [24]. While PUFAs are an important type of fatty acid, our human diets can be varied and contain multiple other types of lipids. Little is known about the effects of other dietary lipids on MeHg toxicity. As high-total-fat, high-saturatedfat, and high-cholesterol diets are all implicated in the development of metabolic syndrome, we were interested in whether changing the bacterial strain fed to *C. elegans* would affect the worm's response to MeHg. We hypothesized that high-fat diets would synergize the metabolic dysfunction caused by MeHg exposure. To test this hypothesis, we exposed worms to MeHg and fed them one of four test *E. coli* diets: the standard diet strain (OP50), HB101 or HT115 (two diets previously shown to cause lower triglyceride accumulation than OP50), or a high-cholesterol diet (OP50 grown on plates containing twice the standard concentration of cholesterol). We then tested for MeHg lethality and Hg accumulation. We determined that the low-fat diets were more protective against MeHg lethality than OP50 and the high-cholesterol diet despite equivalent Hg accumulation. We then examined intracellular triglyceride content and lipid accumulation, as well as pro-adipogenic gene expression in worms exposed to MeHg and fed the test diets. As feeding in *C. elegans* is linked to specific neurobehavior, we assessed feeding and locomotor behaviors controlled by the dopaminergic, serotonergic, and glutamatergic neurotransmitter systems in worms exposed to MeHg and fed the test diets. Finally, as oxidative stress is an important determinant in neurotoxicity and metabolic toxicity, we measured reactive oxygen species (ROS) levels, protein carbonyl content, glutathione content, and antioxidant response element activation following MeHg exposure and test diet feeding.

### **2. Materials and Methods**

### *2.1. Reagents*

Unless otherwise stated, all reagents were obtained from Sigma-Aldrich (St. Louis, MO, USA). Primers used in this study included *tba-1* (F: AGACCAACAAGCCGATGGAG, R: TCCAGTGCGGATCTCATCAAC), *cebp-1* (F: CACTGACATGCCGAACAACG, R: AGAGAGTCTTGTCTTGCGAAGG, *sbp-1* (F: GGCGGCGAAGATTGTGATTC, R: CACTGACATGCCGAACAACG), *fat-6* (F: AGAGGAGAGCAAGAAGATCCCA, R: TCACGGTTTGCCATTTTGCC), and *vit-2* (F: TGATGAGTCCACCAACGAGTTC, R: TTGCTCCTCGTCTCTCTCGT).

### *2.2. C. elegans Strains and Worm Maintenance*

*C. elegans* strains were maintained at 20 ◦C on Nematode Growth Medium (NGM) plates seeded with either *Escherichia coli* strains OP50, HT115, or HB101, as previously described [25]. Additionally, worms were maintained on a 2x cholesterol NGM plate (10 mg/mL cholesterol) seeded with OP50. *C. elegans* are cholesterol auxotrophs; studies have shown that above 5 mg/mL cholesterol levels are high in the nematodes [26]. For the majority of the study, wild-type N2 worms were used. We also used the VP596 (dvls19[pAF15(gst-4::GFP [green fluorescent protein]::NLS)];vsls33[dop-3::RFP (red fluorescent protein)] strain to measure antioxidant response element activity. Both strains were obtained from the Caenorhabditis Genetic Center (CGC; University of Minnesota). The bleaching method was used to harvest eggs for synchronous L1 populations, as previously described [27]. Briefly, embryos were isolated from gravid worms using a bleaching solution (1% NaOCl and 0.25 M NaOH) followed by a sucrose gradient to segregate eggs from worm and bacterial debris. Synchronized L1 worms were treated with MeHg for 30 min in M9 liquid buffer at 25 ◦C on a tube rotator, and then plated on the NGM plates seeded with the differing *E. coli* diets. We have previously shown that these concentrations are below the LD50 for MeHg in *C. elegans* and correlate to concentrations of MeHg in the worm that are below the US EPA reference dose of 0.1 μg/kg/d [28,29].

### *2.3. Dose–Response Survival Curves*

The lethal dose 50% (LD50) of MeHg for N2 *C. elegans* strains fed differing diets was determined by treating 5000 synchronized L1 worms with doses ranging from 1 to 200 μM MeHg for 30 min in M9 liquid buffer at 25 ◦C on a tube rotator. All exposures were carried out in triplicate and repeated 5 times. After treatment, worms were washed 3 times with M9 buffer, transferred to OP50-, HT115-, or HB101-seeded NGM plates or OP50-seeded 2x cholesterol NGM plates, and manually counted for lethality 24 h after MeHg treatment.

### *2.4. Mercury Quantification*

Inductively coupled mass spectrometry (ICP-MS, Agilent 8800 ICP-QQQ) was used to measure intraworm concentrations of Hg. A total of 50,000 worms per sample were treated with MeHg and then fed for 48 h on one of the four test diets before washing with M9 and flash-freezing in liquid nitrogen. The samples were then sonicated. After centrifugation, an aliquot of the supernatant was used to measure protein concentration via the BCA assay. The rest of the sample was digested in the microwave with 1.6 mL bidest H2O, 250 μL HNO3 suprapur® and 250 μL HCl suprapur®. Hg content was measured with No gas mode ICP-MS. Rhodium (0.01 μg/L) was used as internal standard. The calibration was prepared in 10% HNO3 suprapur® and 10% HCl suprapur® using a concentration range of 1–300 ng/L. The washout solution contained 1 ppm gold in 5% HNO3 and 5% HCl. The content of Hg was calculated by dividing total Hg by total protein (ng Hg/mg protein).

### *2.5. Triglyceride Quantification*

The EnzychromTM triglyceride quantification kit (BioAssay Systems, Hayward, CA, USA) was used to measure total intracellular triglycerides. Following the MeHg treatment, 200,000 worms were fed the test diets for 48 h and were homogenized in triglyceride assay buffer. Extracts were incubated for 30 min at room temperature with the triglyceride assay reagent mix, and absorbency (optical density: 570 nm) was read. Data are expressed as mmol triglycerides/μg protein.

### *2.6. Nile Red Staining*

Previously, we have shown that fat storage sites are increased by MeHg through two methods, BODIPY 493/503 and Nile Red [10]. As the Nile Red method amends itself to screening multiple treatment groups, we chose to quantify fat storage sites using this method. Twenty thousand L1 worms were incubated with MeHg, washed, and were transferred to OP50-, HT115-, or HB101-seeded NGM plates or OP50-seeded 2x cholesterol NGM plates. Seventy-two h after treatment, worms were washed off the plates and were fixed for Nile Red staining, as previously described [30]. One thousand worms were first washed with 0.1% triton in PBS, and then fixed in 40% isopropanol for 3 min. Fixed worms were next incubated with 3 μg/mL Nile Red in 40% isopropanol for 30 min followed by another M9 wash step to remove excess dye. Worms were loaded onto a 96-well plate and Nile Red fluorescence was read at excitation 560 nm, emission 590 nm. Data were normalized to worm number and protein levels.

### *2.7. RNA Isolation and Real-Time qPCR Gene Expression*

RNA from 20,000 worms per treatment was isolated using Trizol solution followed by chloroform extraction. cDNA was then synthesized from 1 mg of total RNA using the Appled Biosystems' High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific). Real-time PCR analysis was then performed using PerfeCTa SYBR Green FastMix (QuantaBio, Beverly, MA, USA).

### *2.8. Feeding Behavioral Analysis*

L1 worms were seeded on OP50-, HT115-, or HB101-spread NGM plates, or OP50 spread 2x cholesterol NGM plates following MeHg treatment and were assessed 72 h post-treatment for behaviors associated with nematode feeding; these include pharyngeal pumping, locomotion, and the basal slowing response. For pharyngeal pumping, 10 worms were transferred to fresh NGM/2x cholesterol NGM plates spread with the corresponding *E. coli* diet and the number of pharynx pumps was counted for 30 s. Locomotion was assessed by the body-bend assay: worms were plated on an unseeded NGM/2x cholesterol NGM plate and scored for the number of forward-directed body-bends during a 30 s timespan. The basal slowing response assay was used to measure dopamine-dependent behavior that mediates the worm's slowing movement to consume food. The basal slowing response assay was performed as previously described [31]. The number of forward-

directed body-bends was scored for worms placed either on NGM or 2x cholesterol NGM plates seeded or unseeded with OP50, HB101, or HT115 *E. coli*. For all behavioral assays, 2x cholesterol plates were only used for experiments where worms were fed OP50 grown on 2x cholesterol plates. These data are presented as the change in body-bends, calculated by subtracting the number of body-bends of worms plated on *E. coli*-seeded plates from the number of body-bends of worms plated on unseeded plates.

### *2.9. Glutathione Quantification*

The 5,5- -dithiobis-2-nitrobenzoic acid–GSH disulfide reductase recycling method was used to measure total intracellular glutathione (GSH) levels, as previously described [32] in whole worm extracts from 30,000 worms.

### *2.10. Intracellular Reactive Oxygen Species Determination*

Intracellular reactive oxygen species (ROS) were measured using 2,7-dichlorodihydrofluorescein diacetate (DCFD), as previously described [33]. Briefly, 20 worms treated with MeHg and fed on the test diet for 72 h were loaded onto a black 96-well plate and treated with 25 μM DCFDA. Green fluorescence (excitation 490 nm, emission 520 nm) was read immediately and subsequently every 30 min for 6 h.

### *2.11. Protein Oxidation Quantification*

In this study, 2,4-dinitrophenylhydrazine (DNPH) labeling was used to measure protein carbonylation, a type of protein oxidation, as previously described. Using Yasuda et al.'s method [34], 50,000 treated and test-diet-fed worms were sonicated in 5 mM EDTA with protease inhibitors. Protein was precipitated out of solution using 20% trichloroacetic acid and then incubated with 10 mM DNPH for 1 h. After excess DNPH was washed off, samples were suspended in 6 M guanidine hydrochloride, loaded onto a 96-well plate, and absorbance was read at 380 nm. Concentration of oxidized protein was calculated using Beer–Lambert's law (molar absorptivity coefficient of DNPH is 21 mM<sup>−</sup>1cm<sup>−</sup>1). Data were normalized to total protein concentration.

### *2.12. Oxidative Stress Reporter Assay*

Activation of the antioxidant response element was measured using the VP596 strain, which expresses GFP under the control of the promoter for the SKN-1 target GSH S transferase 4 (*gst-4*). SKN-1, the worm homolog of nuclear factor (erythroid-derived-2) like 2 (Nrf2), is a transcription factor that binds the antioxidant response element and transcribes antioxidant genes in response to environmental insults. VP596 worms also express RFP under the *dop-3* promoter. L1 VP596 worms were treated with MeHg (10 or 20 μM) for 30 min, washed, and transferred to agar plates to be maintained for 72 h on the different diets. Worms were then washed off the plates, loaded onto a 96-well plate, and levels of RFP and GFP florescence were measured (RFP: excitation 544 nm, emission 590 nm; and GFP: excitation 485 nm, emission 520 nm). Antioxidant response element activity was represented as GFP florescence divided by RFP florescence (normalization to worm number).

### *2.13. Statistics*

Statistical analyses were performed using Prism 8 software (Graphpad, San Diego, CA, USA). Statistical analysis of significance was carried out either by Student's *t*-test of LD50 (Figure 1) or two-way analysis of variance (ANOVA) followed by Tukey's multiple comparisons test. Values of *p* < 0.05 were considered statistically significant.

**Figure 1.** Diet affects MeHg toxicity. N2 worms were treated with increasing concentrations of MeHg for 30 min and then transferred to NGM plates seeded with either OP50 or (**A**) HB101, (**B**) HT115, or (**C**) OP50 supplemented with cholesterol. Dose–response survival curves were generated and LD50 values were calculated from five independent experiments. \*\*\* *p* < 0.001 as compared with N2 MeHg-treated worms fed OP50.

### **3. Results**

### *3.1. Bacterial Diet Affects MeHg Toxicity*

As diet is a major environmental factor in health and disease development, we examined whether altering the strain of *E. coli* fed to wild-type N2 worms would affect MeHg toxicity. The standard *E. coli* diet used in nematode culture is OP50. This strain was originally selected due to its ability to form a thin, transparent monolayer, allowing for ease of visualization of the worms under a microscope [35]. How well the OP50 strain emulates the nutrition that a *C. elegans* worm would receive in the wild has not been accurately determined. In addition to the standard OP50 diet, we selected two diets previously shown to be lower in dietary lipids: HB101 and HT115 [36]. In comparison to OP50, worms fed HB101 had ~20% fewer free fatty acids and 50% fewer triglycerides [36]. Furthermore, the fatty acid content of triglycerides in worms fed HB101 contained ~50% fewer branched fatty acids and significantly increased the monounsaturated fatty acid percentage in total fatty acids than worms fed OP50 [36]. In comparison to worms fed OP50, worms fed HT115 had ~50% less triglycerides but did not have significant differences in total free fatty acid content or fatty acid content [36]. Similar to HB101, worms fed HT115 had 50% fewer branched-chain fatty acids in their triglycerides as compared to OP50 worms. Feeding worms either HT115 or HB101 has no effect on the mean lifespan of *C. elegans*, but does result in decreased basal fat storage of dietary lipids [36]. Lastly, we created a high-cholesterol diet by feeding OP50 *E. coli* twice the standard cholesterol concentration in NGM plates. We exposed N2 worms to increasing concentrations of MeHg, plated them on the four different diets, and generated dose–response survival curves (Figure 1). Worms fed either HB101 or HT115 were more resistant to the toxic effects of MeHg, exhibiting a right-hand shift in their dose–response curves as compared to N2 fed OP50 (LD50s of 30.21 and 29.51 μM for HB101 and HT115, respectively, as compared to 20.43 μM for OP50). In contrast, worms fed the 2x cholesterol OP50 diet were more sensitive to MeHg, showing a left-hand shift in their dose–response curves, as compared to N2 fed OP50 (LD50 of 8.91 μM). These data suggest that a bacterial diet affects the toxicity of MeHg in nematodes.

### *3.2. Diet Did Not Alter Mercury Accumulation*

Dietary components have been shown to affect the accumulation of MeHg; for example, the amino acid methionine competes with MeHg for passage through the large amino acid transporter into cells [12]. We therefore were interested in whether the differential toxicity to MeHg among the worms fed the four diets was due to differential accumulation of MeHg in the worms. N2 worms were treated with either 10 or 20 μM MeHg and plated for 48 h on NGM plates that contained one of the four test diets. Levels of Hg in the worms were quantified by inductively coupled mass spectrometry (ICPMS), as previously described [37]. N2 worms fed either the HB101, HT115, or 2x cholesterol OP50 diet accumu-

lated similar levels of Hg following the 10 or 20 μM MeHg treatments as compared to the worms fed OP50 and treated with 10 or 20 μM MeHg (Figure 2). Worms treated with 10 μM MeHg and fed HT115 appeared to have lower levels of Hg accumulation than worms fed OP50 and treated with MeHg; however, this trend was not statistically significant. This suggests that there was no difference in transport, accumulation, or elimination between the worms fed the four diets that could account for the variance in the dose–response curves seen in Figure 1.

**Figure 2.** Bacterial diet does not affect MeHg content in worms. Hg content was measured by ICP-MS in worms fed OP50, HT115, HB101, or OP50+ cholesterol diets 48 h after MeHg treatment. Hg levels are expressed as pg Hg/μg protein. Data represent four independent experiments. \* *p* < 0.05, \*\* *p* < 0.01 as compared to untreated, OP50-fed control.

### *3.3. Bacterial Diet Altered Lipid Accumulation in Response to MeHg*

We have previously shown that MeHg increases the triglyceride content of N2 worms fed OP50 [10]. As HB101, HT115, and 2x cholesterol OP50 contained different lipid profiles to OP50, we were interested in whether the worms would accumulate lipids in response to MeHg at similar levels. L1 N2 were treated with 10 or 20 μM MeHg for 30 min, and were allowed to feed on one of the four test diets and mature for 48 h before extraction of triglycerides. MeHg dose-dependently increased the total triglyceride content in N2 worms (Figure 3A–C). Triglyceride levels in the worms fed HB101 or HT115 were significantly lower than those of the N2 worms treated with 20 μM (Figure 3A,B), while worms fed 2x cholesterol OP50 contained significantly more triglycerides in response to 10 and 20 μM MeHg than the worms fed OP50 (Figure 3C).

**Figure 3.** MeHg increases triglyceride content in worms fed a high-fat, but not low-fat, diet. Total triglycerides were measured in lysates from N2 worms fed OP50 or (**A**) HB101, (**B**) HT115, or (**C**) OP50 supplemented with cholesterol 48 h after MeHg treatment. Data are expressed as mean triglycerides mmol/μg protein ± SEM. All data are representative of five independent experiments. \* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001 as compared to untreated, OP50-fed control. Horizontal bars represent comparisons between Hg-treated worms fed different diets.

We next assessed whether a bacterial diet would affect intracellular lipid storage sites. N2 worms were treated with MeHg, fed one of the four test diets, and, 72 h after exposure, lipid storage sites were stained with Nile Red, as previously described [30]. Both 10 and 20 μM MeHg treatments increased lipid storage in N2 worms (Figure 4A–C). Worms that were fed either the HB101 or HT115 diets showed decreased levels of lipid storage sites in response to MeHg than worms fed OP50 (Figure 4A,B). However, worms fed OP50 with 2x cholesterol showed increased lipid storage sites in response to MeHg compared to worms fed OP50 with standard cholesterol levels (Figure 4C). Taken together these data suggest that lipid accumulation in response to MeHg is altered by the lipid content of the diet.

**Figure 4.** MeHg-induced fat accumulation in worms fed high-fat, but not low-fat, diets. Worms were treated with MeHg for 30 min and placed on NGM plates containing OP50 or (**A**) HB101, (**B**) HT115, or (**C**) OP50 supplemented with cholesterol. Then, 72 h after treatment, worms were fixed, stained with Nile Red, and fluorescence was measured. Data represent mean Nile Red fluorescence normalized to worm number and protein content ± SEM from five independent experiments. \* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001 as compared with untreated, OP50-fed control. Horizontal bars represent comparisons between Hg-treated worms fed different diets.

### *3.4. MeHg-Induced Pro-Adipogenic Gene Transcription Is Diet-Dependent*

We have previously shown that MeHg induces the expression of several genes involved in lipid accumulation and metabolic disease [10]. MeHg increases the expression of *cebp-1* (worm homolog of C/EBP) in worms fed OP50. C/EBP is a regulator of adipocyte differentiation, hyperplasia (increase in adipocyte cell numbers), and hypertrophy (adipocyte cell size) [38,39]. MeHg exposure increased the expression of *cebp-1* in N2 worms fed OP50 72 h post-treatment as compared to untreated controls (Figure 5A). Worms fed with either HB101 or HT115 showed significantly lower expression of *cebp-1* than worms fed OP50. However, *cebp-1* expression levels were not significantly different between the worms fed OP50 or OP50 with 2x cholesterol, suggesting that *cebp-1* induction in *C. elegans* is not dependent on the dietary cholesterol level.

Working in concert with C/EBP to induce adipocyte differentiation in mammals is sterol response element binding protein (SREBP, *sbp-1* in worms) [38,39]. We have previously shown that *sbp-1* expression is increased 24 h after MeHg treatment in worms fed OP50. Here, we report that *sbp-1* levels in response to 10 or 20 μM MeHg remain elevated 72 h post-exposure in OP50-fed worms (Figure 5B). The expression of *sbp-1* followed a similar trend as *cebp-1* in the worms fed HT115 and HB101. These worms had significantly lower expression of *sbp-1* in response to MeHg than the OP50-fed worms. In mammalian systems, the SREBP transcription factor is regulated by dietary cholesterol levels; when there is low cholesterol, the transcription factor is active; however, when there are high levels of cholesterol in cells, SREBP is degraded and the transcription factor is inactivated [40]. Similar regulatory mechanisms are present in worms. Our data show that worms fed OP50 grown on 2x cholesterol plates had no induction of the *sbp-1* gene in response to MeHg.

**Figure 5.** MeHg-induced pro-adipogenic gene expression is dependent on diet. Worms were treated with MeHg for 30 min and placed on NGM plates containing OP50, HB101, HT115, or OP50 supplemented with cholesterol. Then, 72 h after treatment, levels of (**A**) *cebp-1* (ortholog of human C/EBP), (**B**) *sbp-1* (ortholog of human SREBP-1), (**C**) *fat-6* (ortholog to glycerol-3-phosphate acyltransferase), (**D**) *vit-2* were measured by quantitative PCR and normalized to *tba-1* housekeeping gene. Data are expressed as mean relative expression ± SEM from five independent experiments. \* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001 as compared to untreated OP50-fed control. Horizontal bars represent comparisons between Hg-treated worms fed different diets.

Concurrent with the upregulation of adipogenic transcription factor *sbp-1*, MeHg increased the expression of an *sbp-1*-responsive gene, *fat-6*, the worm ortholog of stearoyl-CoA desaturase 1 (SCD). SCD is the rate-limiting step in the formation of monounsaturated fatty acids and triglycerides. Due to its critical roles in in obesity and insulin resistance, SCD is emerging as a potential therapeutic target for these conditions [41]. The transcription of *fat-6* is controlled by multiple transcription factors shown to be affected by MeHg, including *nhr-49* and *sbp-1* [42,43]. As MeHg increases *sbp-1* expression, it is expected that *sbp-1* target genes, such as *fat-6* expression, would also change. Worms treated with MeHg and fed OP50 had increased expression of *sbp-1* (Figure 5B) and increased expression of *fat-6* (Figure 5C). Likewise, worms fed HB101 or HT115 after MeHg treatment had decreased *sbp-1* and *fat-6* expression as compared to OP50-fed worms treated with MeHg. Furthermore, worms fed OP50 raised on NGM plates containing 2x cholesterol did not display *sbp-1* expression or *fat-6* expression, further confirming the relationship of cholesterol levels and *sbp-1* activity in response to MeHg in worms.

Finally, we examined whether diet affected lipid transport proteins in response to MeHg in *C. elegans*. Vitellogenins (*vit-1*–*vit-6*) are yolk proteins with homology to human

apolipoprotein B-100 [44] that deliver cholesterol to oocytes through a receptor-mediated endocytosis mechanism mediated by RME-2, a member of the LDL receptor superfamily. We have previously shown that the levels of *vit-2* were increased by MeHg treatment 24 h post-exposure in N2 worms fed OP50 [10]. In Figure 5D, *vit-2* is increased by MeHg 72 h post-exposure in OP50-fed worms. In worms fed HB101 or HT115, MeHg did not induce *vit-2* expression. Lastly, worms fed OP50 grown on 2x cholesterol NGM plates had increased expression of *vit-2* compared to MeHg-treated worms fed OP50 grown on standard NGM plates. Overall, these data suggest that the different diets affected lipid accumulation in the worms in response to MeHg, which were accommodated by a compensatory modulation of lipid binding and transport proteins.

### *3.5. Feeding Behavior in Response to MeHg Is Dependent on Diet*

Feeding in *C. elegans* is linked to specific neuronal activity. We previously showed that MeHg, a known neurotoxin, increased feeding in worms as well as decreased locomotion and dopaminergic behavior [10]. As we noted differences in fat accumulation between worms fed the four test diets, we were interested in whether there were differences in feeding behaviors. Worms were treated with MeHg and fed for 72 h with either OP50, HT115, or HB101 spread on NGM plates or OP50 spread on 2x cholesterol NGM plates prior to behavioral analyses. Food consumption was measured by the pharynx pump assay. MeHg increased food consumption in worms fed OP50 seeded on standard NGM plates (Figure 6). MeHg did not increase the rate of feeding of worms fed the HB101 or HT115 diets (Figure 6A,B). There was no statistically significant effect of NGM plate cholesterol content on MeHg-induced food consumption; worms fed OP50 on either the standard or 2x cholesterol NGM plates had the same increase in pharynx pumps in response to MeHg (Figure 6C).

**Figure 6.** MeHg increases feeding behavior in worms fed a high-fat diet, but not a low-fat diet. Worms were treated with MeHg for 30 min and placed on NGM plates containing OP50 or (**A**) HB101, (**B**) HT115, or (**C**) OP50 supplemented with cholesterol. Then, 72 h after treatment, food consumption was analyzed by the pharynx pump assay. Data are expressed as means ± SEM from nine independent experiments. \*\* *p* < 0.01, \*\*\* *p* < 0.001 as compared with untreated, OP50-fed control. Horizontal bars represent comparisons between Hg-treated worms fed different diets.

Locomotion and dopaminergic-dependent behavior were also investigated in MeHgtreated worms 72 h following feeding on the four test diets. Previously, we showed that MeHg decreases locomotion rates by measuring forward-directed body-bends [10]. In comparing the four test diets, there was no significant difference in the rates of locomotion in response to MeHg treatment in worms fed either HB101, HT115, OP50, or OP50 on 2x cholesterol NGM plates (Figure 7A). We next measured the change in body-bends on bacteria vs. off bacteria (basal slowing rate, BSR). The BSR is a direct measure of dopaminergic function, as worms deficient in dopamine production (*cat-2* mutants) have no difference in the rates of locomotion on NGM plates with or without a bacterial food source [31]. Healthy N2 worms with functioning dopaminergic systems slow their locomotion on NGM plates spread with bacteria as compared to NGM plates unseeded with bacteria. MeHg is known to be toxic to dopaminergic neurons in mammals and in

*C. elegans* [28,29,45,46]; we were therefore curious as to whether changing the worm's diet might protect the dopaminergic neurons from MeHg-induced dysfunction. BSR was measured in worms treated with MeHg and fed one of the four test diets. MeHg decreased the BSR in worms fed either of the four test diets, and there was no statistical difference in the BSR between worms fed the HB101, HT115, or 2x cholesterol diets as compared to the standard OP50 diet (Figure 7B). This suggests that while MeHg damages the dopaminergic functioning in worms, dopaminergic behavior is not influenced by the bacterial diet strain or cholesterol level.

**Figure 7.** Locomotive and dopaminergic function in response to MeHg are not affected by diet. (**A**) Locomotion behavioral analysis was performed 72 h after MeHg treatment and feeding on the test diets. (**B**) Dopaminergic behavior was assessed by the basal slowing response (BSR) performed 72 h after MeHg treatment and feeding on the test diets. \* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001 as compared with untreated, OP50-fed control.

### *3.6. Bacterial Diet Improves Measures of Oxidative Stress in Response to MeHg*

Oxidative stress is a hallmark of MeHg exposure and can drive neurotoxicity and metabolic alterations. Dietary components can quench ROS [47–50]. We therefore investigated whether altering the bacterial diet fed to worms could prevent oxidative stress derived from MeHg treatment in *C. elegans*. Intracellular ROS were measured in worms treated with MeHg and fed either HB101, HT115, or OP50 on standard NGM plates or OP50 on 2x cholesterol NGM plates by DCFDA staining. MeHg increased intracellular ROS in worms fed OP50 on standard NGM plates (Figure 8A). Worms fed OP50 seeded on 2x cholesterol plates had exacerbated ROS levels in response to MeHg. Worms fed either the HB101 or HT115 diet had significantly less MeHg-induced ROS generation. These data suggest that the dietary components in the different *E. coli* strains produced differential oxidative stress in response to MeHg. ROS damages cellular biomacromolecules, leading to lipid peroxidation and subsequent carbonyl protein adduct formation on cysteine, lysine, and histidine amino acids through Michael addition chemistry. Levels of oxidized proteins were measured using a DNPH colorimetric assay that quantified carbonyl adducts on proteins in samples derived from lysates of N2 worms treated with 20 μM MeHg and fed one of the four test diets. MeHg significantly increased the protein carbonyl content in worms fed OP50 seeded on standard NGM plates, which was significantly decreased in worms fed either HB101 or HT115 (Figure 8B). In contrast, worms fed OP50 seeded on 2x cholesterol NGM plates had exacerbated protein carbonyl content as compared to worms fed OP50 seeded on standard NGM plates.

We next investigated the worms' ability to mount an antioxidant response to MeHg following feeding with the test diets. GSH is the main intracellular thiol that is responsible for maintaining the redox environment of the cell. MeHg readily binds free thiols, such as those present on GSH. As previously observed [28], treatment with 10 or 20 μM MeHg led to a 20% decrease in total GSH levels in worms fed OP50 seeded on standard NGM plates (Figure 8C). Worms fed HB101 had significantly higher basal levels of GSH than worms fed OP50. MeHg treatments decreased the levels of GSH in HB101-fed worms as compared to untreated worms fed HB101; however, the levels of GSH in MeHg-treated HB101 fed worms were significantly higher than those in OP50-fed MeHg-treated worms. Worms fed HT115 showed a minimal decrease in GSH in response to MeHg; however, in comparing worms treated with MeHg, the HT115-fed worms contained more GSH than the OP50-fed worms. Similar to the oxidized protein and intracellular ROS data, worms fed OP50 seeded on 2x cholesterol plates had exacerbated GSH loss in comparison to worms fed OP50 seeded on standard NGM plates.

**Figure 8.** Diet affects MeHg-induced oxidative stress. Worms were treated with MeHg for 30 min and placed on agar plates spread with either OP50, HB101, Ht115, or OP50 supplemented with cholesterol. Measures of oxidative stress were assessed 24 h after MeHg treatment. (**A**) ROS levels were measured through DCFDA fluorescence. Data are expressed as mean fluorescence ± SEM for 6 independent experiments (**B**) Protein carbonyl levels were measured and normalized to protein content. Data are expressed as means ± SEM from 5 independent experiments. (**C**) Total GSH levels were measured and normalized to protein content. Data are expressed as mean ± SEM from five independent experiments. (**D**) Quantification of GFP fluorescence of VP596 transgenic worms expressing GFP under the *gst-4* promoter. Data are expressed as means fluorescence ± SEM from 5 independent experiments. \* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001 as compared with untreated, OP50-fed control. Horizontal bars represent comparisons between Hg-treated worms fed different diets.

> Finally, we examined the effect of the different diets on the ability of the worms to generate antioxidant defense proteins. Phase II metabolic genes and antioxidant defense enzymes are regulated by the activity of the Nrf2 (Nuclear factor erythroid 2-related factor 2) (SKN-1 in *C. elegans*) transcription factor. MeHg is known to induce Nrf2 in cell culture and in nematodes [51,52]. We used the VP596 strain, which expresses GFP under the control of the promoter for the SKN-1 target *gst-4*. MeHg treatment significantly increased the amount of GFP fluorescence in worms fed OP50 seeded on standard NGM plates (Figure 8D), indicating increased SKN-1 activity. Worms fed either HB101 or HT115 had significantly less GFP fluorescence in response to MeHg treatment than the OP50-fed worms. This suggested that there was less oxidative stress, and therefore less SKN-1 activity, in HB101- and HT115-fed worms than in the OP50-fed worms. In contrast, worms fed OP50 seeded on 2x cholesterol NGM plates had significantly more GFP fluorescence in response to MeHg than worms fed OP50 seeded on standard OP50, suggesting the presence of greater oxidative stress and SKN-1 activity.

### **4. Discussion**

Herein, we demonstrate for the first time that the toxicity of MeHg in *C. elegans* is dependent on the strain of *E. coli* used as a food source. MeHg is a known neurotoxic agent that has long been understood to cause neurological changes and is emerging as a player in metabolic diseases. *C. elegans* is a useful model organism to study the effects of the metabolic changes induced by MeHg because of its evolutionarily conserved fat and sugar metabolic pathways [53]. We have previously shown that MeHg causes metabolic alterations in *C. elegans* that lead to decreased nicotinamide adenine dinucleotide (NAD+) cofactor levels, mitochondrial dysfunction, and oxidative stress [28]. We also found that MeHg increases the transcription of *cepb-1* (ortholog to human C/EBP), *nhr-49* (ortholog to human peroxisome proliferator activated receptor gamma, PPARγ), and *sbp-1* (ortholog to human sterol response element binding protein-1, SREBP-1), pro-adipogenic transcription factors implicated in MS, as well as a number of other genes involved in lipid synthesis and transport [10]. All of these findings were under the context that the worms were consuming the standard OP50 *E. coli* diet. It was recently shown that feeding *C. elegans* dehydrated dead OP50 significantly decreased the susceptibility of worms to MeHg [54]. These data suggest that the diet fed to *C. elegans* is an important determinant of its susceptibility to the toxic effects of MeHg.

MeHg enters the human body primarily through fish consumption. For decades, studies of populations in the Seychelles, Faroe Islands, and Arctic Canada have followed cohorts of individuals exposed to MeHg in their diet and have produced, at times, conflicting data [55–57]. While there are multiple explanations for these discrepancies, from the types of fish eaten, to co-contaminants (such as polychlorinated biphenyls, PCBs) and/or genetic polymorphisms present in the populations, background diet composition has not been fully taken into consideration in these comparisons. Our data show in worms that the strain of *E. coli* can drastically alter the toxicity of MeHg independently of the amount of MeHg that accumulates in the worm. The strains HB101 and HT115, which are lower in triglycerides and free fatty acids than OP50 [36], conferred protection to worms from MeHg lethality, lower fat accumulation and adipogenic gene transcription, and decreased levels of oxidative stress than worms fed OP50. Conversely, worms fed OP50 supplemented with excess cholesterol were more susceptible to MeHg and had increased lipid accumulation and oxidative stress as compared to worms fed OP50 with a standard concentration of cholesterol in the NGM. Recent data have shown that *C. elegans* in the wild eat a varied diet comprising bacteria not only in the genus *Escherichia*, but also in the genera *Sphingomonas*, *Xanthomonas*, and *Methylobacterium* [58]. These diets confer differences in lifespan, development, reproduction, and gene expression [58]. It remains to be determined how feeding any of these non-*E. coli* species to worms might affect the toxicity of MeHg.

Our data examining lipid accumulation and MeHg toxicity in the context of higherand lower-fat diets in *C. elegans* are in agreement with previous findings in rodents and humans. KK-Ay type 2 diabetic mice, which have higher body fat than C57BL/6J mice, had lower blood clearance of MeHg and increased neurological damage compared to C57BL/6J mice [59,60]. Likewise, high-fat diets and MeHg were shown to result in similar lipid and cholesterol accumulation and steatosis in the liver [61]. In a toxicokinetic study, Rowland et al. fed mice either a standard diet or a synthetic diet that was high in protein and low in fat. The mice fed the synthetic diet accumulated less Hg in their body than the standard-pellet-fed mice [62]. However, it was unclear whether the effect of this synthetic diet was due to the low fat, the high protein, or to both factors. Since Hg interacts with multiple amino acids, such as cysteine, or is antagonized by methionine, the protective effect that Rowland et al. observed may have been due to the amino acid content of the synthetic diet.

MeHg has been shown to increase cholesterol levels in humans and rodents. MeHg has been shown to inhibit paraoxonase activity, leading to increased low-density lipoproteins (LDL) in an Inuit population that regularly eats fish [63]. Additionally, chronic exposure to MeHg in wild-type and LDL receptor knockout mice causes hypercholesterolemia [64]. Our data support the notion that a diet rich in cholesterol potentiates the toxic effects of MeHg on lethality and lipid accumulation. While we did not specifically measure cholesterol levels in our worms following MeHg exposure and feeding on the 2x diet, our gene expression analysis showed that high cholesterol levels were indeed achieved

in the worms, allowing for the inhibition of *sbp-1* transcription. SREBP1 (*sbp-1* in worms) is a transcription factor that upregulates steroid and cholesterol biosynthetic genes but is inactivated and degraded under high cholesterol levels [40].

The central nervous system plays an important role in sensing nutrients and integrating hormonal signals from the gastrointestinal tract and adipocytes to regulate caloric intake and energy expenditure [65]. In humans, the hypothalamic–pituitary axis (HPA) integrates hormonal signals including leptin, insulin, ghrelin, and adiponectin. The HPA is especially vulnerable to neurotoxicants, such as MeHg, as the blood–brain barrier is weak in certain areas, such as the arcuate of the hypothalamus, which produces neuropeptide Y [66]. In vitro studies of hypothalamic neuronal cell lines treated with MeHg show increased expression of neuropeptides pro-omiomelanocortin (Pomc) and Agouti-related peptide (Agrp), key regulators of homeostasis [67]. Just as the CNS controls feeding and energy expenditure, diet and nutrients can affect CNS function. Diet-induced obesity and high-fat diets reduce dopamine release and reuptake, leading to disruption of the satiety circuits between nucleus accumbens (NAc) dopamine terminals and projections to the hypothalamus [68,69]. Long-term feeding of high-fat diets in mice depletes dopamine in the NAc, which may contribute to the development of obesity [70]. Conversely, caloric restriction in rats has led to increased dopamine and serotonin levels in the striatum and increased leptin levels in the plasma [71].

The *C. elegans* nervous system shows simplified neuronal control of nutrient sensing. Both humans and nematodes use dopamine, glutamate, and serotonin to control foraging, locomotion, feeding, and nutrient sensing [72]. MeHg disrupts both dopamine and glutamate signaling, while little is known about MeHg's effects on serotonin. *C. elegans* use dopamine signaling to sense food, increase turn frequency when leaving food, and for defecation [31,73,74]. We have previously shown that MeHg decreases dopamine levels and behaviors in *C. elegans* fed OP50, leading to decreased locomotion and deficits in sensing the presence of food [28]. In our present study, we observed that MeHg decreased dopaminergic activity in the worms independently of which bacterial strain or cholesterol concentration was presented during the feeding. This was an unexpected result. Previously, we have shown that supplementing worms with excess nicotinamide adenine nucleotide can prevent dopaminergic damage in response to MeHg in *C. elegans* [28]. Since the low-fat HB101 and HT115 diets were more protective for MeHg lethality and metabolic dysfunction, we hypothesized that there would be less dopaminergic damage. However, MeHg is a well-characterized dopaminergic toxicant; the lipid level of the diet presented to the worm was irrelevant. MeHg caused significant dopaminergic damage that was not prevented.

The physical act of feeding in nematodes is measured by the rate of pharyngeal pumping. Serotonin regulates the pharynx muscles, allowing for food to be drawn through the mouth upon muscle contraction [75]. Serotonergic neurons coordinate the action of the cholinergic MC and glutamatergic M3 motor neurons that directly synapse on pharyngeal muscle cells [76]. Glutamate is released from the M3 neurons and activates glutamategated chloride channel AVR-15 expressed on pm4 and pm5 pharyngeal muscle cells, leading to the modulation of the duration and frequency of pharyngeal pumping [76,77]. Mutants deficient in glutamate signaling lose the ability to terminate action potentials on the pm4 and pm5 cells, resulting in a reduced pumping rate [78]. MeHg disrupts glutamate signaling, leading to an increased glutamate concentration in the synapses and neuronal excitotoxicity [79–81]. Therefore, disruption of glutamatergic signaling by MeHg can negatively affect the rate of pharynx pumping. Previously, we have reported increased feeding in response to MeHg in *C. elegans* fed OP50 [10]. In our present study, worms fed HB101 or HT115 showed no increase in feeding in response to MeHg. This suggests that there may have been less damage by MeHg to the glutamatergic or serotoniergic neurotransmitter systems in HB101- and HT115-fed worms than in worms fed OP50. Decreased feeding, as compared to OP50, may be one of the mechanisms by which the low-fat diets protected the worms from lipid dysregulation in response to MeHg. It is important to note that, basally, different bacterial species and bacterial strains can cause

differential pharyngeal pumping in *C. elegans* [36,58]. Both HB101 and HT115 have been shown in L4 and young adult worms to lead to significantly lower rates of pharyngeal pumping as compared to worms fed OP50 [58,82,83]. Our data show no difference in basal pharynx pump rate between worms fed OP50, HB101, or HT115. This may be due to our use of older worms (adults ~72 h post L1 stage) than in previous reports. Our data also show that the cholesterol concentration in the diet did not affect the pharyngeal pumping rate basally or in response to MeHg. Worms fed OP50 grown on plates with 2x cholesterol concentration had a dose-dependent increase in feeding similar to worms fed OP50 grown on NGM plates with the standard cholesterol concentration.

While the nervous system is a key modulator of nutrient sensing, nutrients themselves actively signal to neurons to regulate feeding behaviors. Nematodes and mammals express neuropeptides that signal to neurons in response to the presence of nutrients. For example, FLP-20 regulates glutamatergic neurons to degrade fat and induce autophagy following starvation [84]. Nutrients send either feedforward or feedback modulation to the neurons that control the pharyngeal pump rate to either increase or decrease feeding [85]. It is unknown how these pathways are affected by MeHg exposure.

MeHg exerts many of its toxic and neurotoxic effects due to the induction of oxidative stress. ROS generation following MeHg exposure can damage membrane lipids, leading to lipid peroxidation, loss of membrane integrity, and dysfunction of neuronal signaling. Likewise, oxidative stress following MeHg exposure can alter gene expression or directly damage enzymes and transport proteins. Dietary factors can also influence ROS levels and oxidative stress. In our study, ROS levels were increased in worms fed OP50 following MeHg treatment and were exacerbated when worms were fed the 2x cholesterol diet. This is in agreement with previous studies demonstrating that high-fat diets are linked to increased oxidative stress, leading to mitochondrial dysfunction and additional ROS production [86,87]. Oxidative stress induced by high-fat diets has been shown to be blocked by dietary factors, such as yogurt, quercetin, geraniin (polyphenol derived from *Nephelium lappaceum* L. fruit rind), and *Terminalia arjuna* extract, to name a few [47–50]. Indeed, in our study, worms fed HB101 or HT115 following MeHg exposure had significantly lower ROS production than the worms fed OP50. ROS can damage lipids, leading to lipid peroxidation, and subsequently oxidizes proteins via a process known as protein carbonylation [88]. We observed that protein carbonylation in our study was consistent with our ROS production data. Worms fed OP50 following MeHg exposure had high levels of protein carbonylation, which was exacerbated by feeding the worms the 2x cholesterol diet. Conversely, worms fed the HB101 or HT115 diets following MeHg exposure had lower levels of protein carbonylation as compared to worms fed OP50. MeHg depletes cellular GSH, leading to a reduced capacity to buffer oxidative stress [89,90]. In *C. elegans*, GSH was significantly decreased in worms fed OP50 or the 2x cholesterol diet following MeHg treatment. Worms fed the HB101 or HT115 diet did not lose intracellular GSH content following MeHg exposure. While this may have been the result of less oxidative stress resulting from MeHg in the HB101- or HT115-fed worms, the bacterial strain fed to the worms cannot be ruled out. While total protein content is not significantly different between OP50, HT115, and HB101 [36], it remains to be determined whether the thiol levels between the three strains are similar. As a final measure of oxidative stress, we measured fluorescence from a GFP reporter strain that fluoresces upon activation of the antioxidant response element. Worms fed the OP50 or 2x cholesterol diet had increased induction of the GFP reporter in response to MeHg, suggesting high levels of antioxidant gene induction. Worms fed HB101 or HT115 following MeHg had significantly lower levels of GFP induction than worms fed OP50, corroborating that there were lower levels of oxidative stress in response to MeHg in worms fed either of these lower-fat bacterial strains.

### **5. Conclusions**

Altogether, our study demonstrates that dietary lipid content and cholesterol content are major determinants in the response of *C. elegans* to MeHg. Worms fed diets low in lipids had reduced triglyceride and lipid accumulation in response to MeHg, ate less food, and experienced less oxidative stress than worms fed the standard OP50 diet that was higher in lipid content,. Conversely, worms fed a diet high in cholesterol had increased triglyceride and lipid accumulation in response to MeHg, and experienced more oxidative stress than worms fed the standard OP50 diet. Diet did not affect certain neurotoxicities in response to MeHg, such as dopaminergic dysfunction; however, diet did affect the rate of feeding. These data suggest that MeHg-induced lipid dysregulation and oxidative stress is influenced by dietary factors, such as triglycerides and cholesterol, leading to metabolic changes characteristic of obesity and metabolic syndrome.

**Author Contributions:** Conceptualization, S.W.C.; methodology, J.L.N.-C., J.B., T.S. and S.W.C.; formal analysis, J.L.N.-C., J.B. and S.W.C.; investigation, N.C., M.M., T.N., B.K., F.I., E.N., J.L.N.- C., J.B., T.S. and S.W.C.; resources, J.L.N.-C., T.S. and S.W.C.; writing—original draft preparation, J.L.N.-C., J.B. and S.W.C.; writing—review and editing, S.W.C.; supervision, J.L.N.-C., J.B. and S.W.C.; project administration, S.W.C.; Funding acquisition, J.B., T.S. and S.W.C. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Husson University School of Pharmacy Research Grant, Husson University Research Fund Program, DFG Research Unit TraceAge (FOR 2558).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.

### **References**


**Sara Gómez-Arnaiz <sup>1</sup> , Rothwelle J. Tate <sup>2</sup> and Mary Helen Grant 1,\***


**Abstract:** Metal-on-metal (MoM) hip implants made of cobalt chromium (CoCr) alloy have shown early failure compared with other bearing materials. A consequence of the abnormal wear produced by these prostheses is elevated levels of cobalt in the blood of patients, which can lead to systemic conditions involving cardiac and neurological symptoms. In order to better understand the implications for patients with these implants, we carried out metal content and RNA-Seq analysis of excised tissue from rats treated intraperitonially for 28 days with low concentrations of cobalt. Cobalt blood levels in dosed rats were found to be similar to those seen in some patients with MoM implants (range: 4–38 μg/L Co in blood). Significant accumulation of cobalt was measured in a range of tissues including kidney, liver, and heart, but also in brain tissue. RNA-Seq analysis of neural tissue revealed that exposure to cobalt induces a transcriptional response in the prefrontal cortex (pref. cortex), cerebellum, and hippocampus. Many of the most up- and downregulated genes appear to correspond to choroid plexus transcripts. These results indicate that the choroid plexus could be the brain tissue most affected by cobalt. More specifically, the differentially expressed genes show a disruption of steroidogenesis and lipid metabolism. Several other transcripts also demonstrate that cobalt induces an immune response. In summary, cobalt exposure induces alterations in the brain transcriptome, more specifically, the choroid plexus, which is in direct contact with neurotoxicants at the blood–cerebrospinal fluid barrier.

**Keywords:** metal-on-metal (MoM) hip implants; cobalt; systemic cobaltism; neurotoxicity; RNA-Seq; RT-qPCR

### **1. Introduction**

Hip arthroplasty procedures successfully lead to the reduction in pain and improved mobility in patients suffering from joint diseases such as osteoarthritis. However, a decade ago, consultants and regulatory bodies reported their concerns over the failures of certain models of metal-on-metal (MoM) hip implants, which resulted in the market withdrawal of the Articular Surface ReplacementTM (ASRTM) hip implant from DePuy Orthopaedics [1]. Recently, these concerns have been extended to all MoM implant models as it has been demonstrated that MoM implants induce adverse reaction to metal debris (ARMD), which includes metallosis, pseudotumours, aseptic lymphocytic vasculitis associated lesion (AL-VAL), and even necrosis [2]. The metal debris is released from the bearing surface, and the taper junction in the case of the total hip replacements (THR), due to wear and corrosion of the metallic parts [3]. Metal ions eventually dissolve and are released into the bloodstream [4]. Co and Cr metal ions in blood are indicative of implant failure and tests for metal ion concentrations should be performed for all patients with MoM implants in the UK according to the MHRA updated guidelines [5]. The European Commission also recommends the monitoring of metal ion levels for MoM implant patients [6].

Neurological conditions thought to be caused by high levels of cobalt ions in blood have been demonstrated in several clinical reports in relation to patients with MoM im-

**Citation:** Gómez-Arnaiz, S.; Tate, R.J.; Grant, M.H. Cobalt Neurotoxicity: Transcriptional Effect of Elevated Cobalt Blood Levels in the Rodent Brain. *Toxics* **2022**, *10*, 59. https://doi.org/10.3390/ toxics10020059

Academic Editors: Richard Ortega and Asuncion Carmona

Received: 30 November 2021 Accepted: 13 January 2022 Published: 28 January 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

plants [7–11]. These involve a range of symptoms such as cognitive decline, memory loss, and mood disturbances, in addition to other visual and auditory issues and peripheral neuropathy. There is little information about the actual effects of cobalt in the peripheral and the central nervous system, and the most appropriate response to diminish patients' symptoms to date is to remove the implant as the source of cobalt ions through revision surgery [8,12,13]. However, the recovery process from cobaltism is not well documented and there is no unified action considered appropriate to treat these patients. Our lack of knowledge on cobalt actions in the body extends further since we also fail to consider the implications for asymptomatic patients with high levels of cobalt in the blood [14].

The effects of cobalt have been extensively researched in vitro with models such as astrocytes [15,16], neurons [17–21], and glial cells [22,23], which are valuable tools to understand the modes of action of cobalt toxicity in the brain. Nevertheless, the nature of in vitro work requires highly controlled experimental conditions, and this can diminish an organism's complexity. Previous toxicogenomic analyses have demonstrated that in vitro systems fail to fully represent relevant processes occurring in rat liver tissues in vivo after exposure to toxic compounds [24]. Molecular and functional events occurring in vivo at the tissue and organ levels could be crucial mechanisms in the physiological response to cobalt. In this sense, transcriptomic applications have already been proven to be effective in toxicology, not only by understanding the mechanisms, but also by finding early end-points for the detection of toxicity and identification of biomarkers [25]. However, most in vivo cobalt studies have focused on the study of reactive oxygen species and the expression of hypoxia markers. Pregnant female rats dosed orally with 350 mg/L delivered pups with impaired levels of antioxidant proteins in the cerebrum and cerebellum [26]. Caltana et al. observed that the direct cortical injection of cobalt led to histological changes consistent with focal ischaemia involving neuronal and astrocyte morphological changes [27]. Another group applied cobalt dust directly into the dura mater [28] and discovered an elevated expression of proteins involved in thyroid transport and regulation of glycolysis. Nevertheless, it is difficult to establish the relevance of these studies for patients with MoM implants due to the high dosage of cobalt used in the animals, the different types of cobalt delivery methods, and the missing information on the resulting cobalt concentrations in blood or plasma. Our research mimicked the conditions that MoM implant patients endure to obtain a better representation of relevant cobalt toxic mechanisms for them.

The aim of this in vivo work was to identify how cobalt accumulated in tissues after long-term systemic circulation, and specifically answer whether cobalt concentration increased in the brain. For that, we carried out comprehensive time- and dose-response experiments to cobalt treatment. We also sought to understand the underlying molecular changes in the brain at the transcriptional level in laboratory rats after prolonged doseresponse exposure. Research on MoM cobalt-induced systemic toxicity is scarce and to our knowledge, there have been few publications addressing neurological manifestations *in vivo*. We hope to generate and test hypotheses through RNA-Sequencing (RNA-Seq) to gain mechanistic insights into the modes of action of cobalt.

### **2. Materials and Methods**

### *2.1. Experimental Animals and Research Design*

Experiments were performed in adult male Sprague Dawley (SD) rats obtained from Charles River (UK). The body weight range at the start of the experiments was 210–280 g. Food and water were provided *ad libitum*. Their body weight and general aspects of health were monitored daily.

Freshly-made cobalt chloride hexahydrate (CoCl2.6H2O; Sigma-Aldrich, Dorset, UK) solutions dissolved in distilled water (dH2O) and dH2O were sterilised through a 0.22 μm syringe-driven filter (Merck Millipore, Watford, UK).

Two separate in vivo experiments were performed successively. The first experiment was a time-response experiment in which a 7-day cobalt treatment was compared against a 28-day treatment at a fixed cobalt dose. Animals were treated daily with either 1 mL/kg

body weight dH2O (controls) or 1 mg/kg body weight (BW) CoCl2 doses injected i.p. (treated groups). The second in vivo experiment was a dose-response experiment. The animals were given 1 mL/kg dH2O i.p. in the case of the control group, or a range of cobalt solutions- 0.1, 0.5, or 1 mg/kg BW i.p. injections for 28 days. Figure 1 depicts the sample size and design information diagrammatically.

**Figure 1.** Design of in vivo time- and dose-response experiments showing group distribution and sample size. All injections, both control and Co-treated rats, were carried out intraperitonially.

### *2.2. Sacrifice of Animals and Tissue Harvest*

At the end of the exposure time, animals were killed by carbon dioxide (CO2) asphyxiation in a CO2 chamber. Each organ or brain part was dissected separately, weighed, and sections of tissue were stored appropriately to preserve its metal content and molecular RNAs. Blood samples were collected through cardiac puncture immediately before death and mixed with 200 μL of heparin (1000 IU/mL diluted 1:10; Sigma-Aldrich, Dorset, UK) to avoid clotting.

### *2.3. Tissue Cobalt Content Measured by ICP-MS*

Quantification of cobalt content in all organs was obtained via inductively coupled plasma mass spectrometry (ICP-MS) analysis. For this, 100 mg of tissue from each organ/area of interest was taken and stored at −80 ◦C until further sample digestion. To obtain a liquid solution of the samples suitable for metal detection, 0.5 mL concentrated nitric acid was used per sample (HNO3; TraceSELECTTM Ultra, Sigma-Aldrich (Fluka), Dorset, UK) followed by 0.25 mL 30% hydrogen peroxide (*w*/*w*) (H2O2; Sigma-Aldrich, Dorset, UK). Each reagent was to left act for 20 min at 100 ◦C in a hot block to ensure the decomposition of the matrix. A quantity of 0.25 mL was transferred together with 9.75 mL ultrapure water into acid-washed tubes to avoid trace metal contamination. Standard dilutions were prepared from Cobalt Standard for atomic absorption spectrometry stock (TraceCERT®, Sigma-Aldrich (Fluka), Dorset, UK). The 1:40 dilution samples were serially introduced to the Agilent 7700× octopole collision ICP-MS system (Agilent Technologies, Cheadle, UK). Scandium or indium was used as the internal standard and data were obtained from the maximum signal with the ICP-MS operating in helium mode.

### *2.4. RNA Extraction*

### 2.4.1. Isolation of RNA

To reduce RNA degradation, 50 or 100 mg from specific segments of brain tissue were dissected and placed in RNase-free tubes with 0.5 mL of RNAlater (Ambion, Life Technologies, Paisley, UK). RNase-free aerosol-resistant filter tips were used for all pipetting steps. Tissues were further cut into thin pieces and completely submerged in the stabilisation solution for overnight storage at 4 ◦C. RNAlater reagent was removed the next day and the tissues were frozen at −80 ◦C until RNA extraction.

Tissue was then resuspended in 1 mL QIAzol lysis reagent (Qiagen, Crawley, UK) and a cone ball steel bead was placed inside the tube. The tissue was disrupted with a horizontal Retsch MM200 Mixer Mill (Retsch GmbH, Haan, Germany) set at 30 Hz for 1 min intervals (×3). The lysate was transferred to another tube with 4 mL of QIAzol lysis reagent and the contents vortexed. The pre-processed samples were further prepared according to the protocol of the RNeasy Plus Universal Midi Kit (Qiagen, Crawley, UK).

### 2.4.2. Quality Check of RNA Samples

RNA purity characterised by the ratio of the absorbances 260/280 nm and 260/230 nm as well as nucleic acid concentration were quantified with a Nanodrop-2000c spectrophotometer (Labtech International, Heathfield, UK). Sample integrity was assessed through microfluidic chip-based analysis (Experion RNA StdSensChip; Bio-Rad, Watford, UK) in the Experion Automated Electrophoresis System (Bio-Rad, Watford, UK) and evaluated with the RNA Quality Indicator (RQI) number [29]. The results of these analyses for pref. cortex, cerebellum, and hippocampus are displayed in Tables S1–S3, respectively, in the Supplementary Materials.

### *2.5. RNA Sequencing (RNA-Seq)*

### 2.5.1. Sample Pooling for RNA-Sequencing

To reduce the RNA-Seq costs, four biological samples for each comparison group were pooled into a single sample. The necessary volume from each sample was adjusted to obtain an RNA quantity of 5–20 μg depending on the brain area yield. The individual sample volumes were pooled into the corresponding RNase-free tubes for each group and the final contents were briefly vortexed.

### 2.5.2. RNA-Sequencing Analysis by BGI

RNA-sequencing of the pooled samples was performed by BGI Tech Solutions (Hong Kong, China) Co. Ltd. using a BGISEQ-500 sequencing platform with depth of 20 million base pairs (Mb) clean reads per sample and a 50 single-end bases (50SE) read length. Filtering of clean reads and their mapping to the UCSC rn6 rat reference genome were carried out by BGI, in addition to the quantification of gene fold-change.

### 2.5.3. Gene Ontology (GO) and KEGG Pathway Enrichment Analysis

Further bioinformatic analyses such as hierarchical clustering or generation of Venn diagrams were performed in MATLAB software (The MathWorks Inc. Natick, MA, USA. Release R2021a Update 5, 9.10.0.1739362). The software tool Cytoscape and its plugin ClueGO were used to conduct the Gene Ontology analysis of the DEGs [30,31]. ClueGO allows for a comparison to different reference ontology sets such as Molecular Function (GO MF; 8 April 2016), Biological Process (GO BP; 8 April 2016), and Cellular Component (GO CC; 8 April 2016), which describe specific gene function and cellular location aspects of gene activity as well as the Kyoto Encyclopaedia of Genes and Genomes (KEGG; 14 June 2016) for the specific enrichment of known pathways. Additionally, other online software was used to create protein–protein interaction (PPI) networks of gene-protein products: STRING (https://string-db.org/) [32] (last accessed: 24 December 2020).

### *2.6. Quantitative Real-Time Polymerase Chain Reaction (RT-qPCR)*

Real-time quantitative PCR (RT-qPCR) was used to obtain the expression of genes of interest in individual samples in order to validate the RNA-Seq results from the pooled samples. All RT-qPCR experiments were performed in accordance with the minimum information for the publication of quantitative real-time PCR experiments (MIQE) guidelines [33]. The checklist for RT-qPCR assays performed with samples from the in vivo experiments is fully displayed in Table S4 in the Supplementary Materials.

The LunaScript RT SuperMix Kit (New England Biolabs, Hitchin, UK) was utilised to generate cDNA for the dose-response experiment samples. To each RT+ or RT- reaction, enough volume of each sample was added to attain 1 μg RNA in each tube. The thermal cycler (Model 480, Perkin Elmer, Warrington, UK) was pre-heated at 25 ◦C, and primers were allowed to anneal for 2 min at this temperature, continuing at 55 ◦C for cDNA synthesis during 10 min, and the final RT inactivation at 95 ◦C for 1 min. After cooling down the samples, the cDNA samples were stored at −20 ◦C until further use for RTqPCR reactions.

The primer sequences were designed to span an intron or exon–exon junction through the NCBI Primer-BLAST tool (https://blast.ncbi.nlm.nih.gov/Blast.cgi) to selectively amplify cDNA (last accessed: 3 March 2020). The primer design criteria followed is displayed in Table S5 in the Supplementary Materials. Oligos were synthesised and purchased from Integrated DNA Technologies (IDT, Leuven, Belgium). The genes with their accession number and primer sequence, amplicon length, and melting temperature are presented in Table S6.

PCR experiments were performed with triplicate RT+ per sample, one no-reverse transcriptase control (RT-) per sample, and one no-template control (NTC) per gene and with the corresponding reference gene controls. The SYBR Green detection method was used for the detection of amplification with the kit PowerUpTM SYBRTM Green Master Mix (Applied Biosystems, Thermo Fisher Scientific, Paisley, UK). The forward and reverse primers (10 pM/μL) and molecular-grade water were mixed with Master Mix (2×). Samples consisting of 1 μL cDNA were pipetted into MicroAmp Fast Optical 96-well reaction plates (Applied Biosystems, Paisley, UK) with the previous mix to a final volume of 20 μL per well. The specific thermal cycling parameters (Table S7) were set according to the optimised PowerUpTM SYBRTM Green Master Mix for fast cycling mode in the StepOnePlus Real-Time PCR system (Applied Biosystems, Paisley, UK). At the end of the cycling process, a melt curve was produced and inspected for the occurrence of primer dimers, misprimes, and possible contamination of genomic DNA.

The method used to calculate the relative fold-change was the comparative CT method [34], with CT being the threshold cycle detected over the 40 run cycles. For gene normalisation, the expression of typical brain reference genes *Ywhaz*, *Tbp*, and *Pes1* was studied. Primer sequences are shown in Table S8. After quantification, the RefFinder web-based tool: https://www.heartcure.com.au/reffinder/ [35] was used to define the most stable reference gene for each tissue (last accessed: 2 March 2020). Full results of RefFinder analyses for each tissue are shown in Figures S1 and S2 in the Supplementary Materials.

### *2.7. Statistics*

Shapiro–Wilk and Levene's tests were used as preliminary methods to evaluate data normality and homogeneity of variances, respectively. Independent samples *t*-test was performed to establish statistical comparisons between the two groups (i.e., control and treatment groups). For statistical comparisons between more groups, the tests selected were one-way analysis of variance (ANOVA) together with the Dunnett's post-hoc test. Statistical significance was declared when *p*-value < 0.05. The statistical software used was IBM SPSS Statistics 25 (IBM Corp. Armonk, NY, USA. Released 2017. IBM SPSS Statistics for Macintosh, version 25.0).

### **3. Results**

### *3.1. Cobalt Accumulates in Organs and in Different Brain Structures*

Figure 2 displays cobalt accumulation in all tissues tested and blood after 7- and 28-day treatments. ICP-MS results reveal that kidney, liver, and heart incorporated the highest Co content among the rats' organs in that order. After 28 days, pref. cortex, cerebellum, and hippocampus assimilated significant amounts of cobalt (*p* < 0.01, compared with the control rats). This was not the case for the 7-day treatment. Co levels detected in blood from the treated rats by ICP-MS were within the range found in MoM patients. After 28 days, Co detected in rat blood was 27.14 ± 2.70 μg/L compared with the average levels of 1–2 μg/L reported in MoM hip resurfacing patients [36]. The maximum cobalt concentration found in a patient with a THR prosthesis was 6521 μg/L [37].

**Figure 2.** Cobalt content in SD male rats' tissues (ng/g) and blood (μg/L) at 7- and 28-days of daily i.p. CoCl2 injection treatment as assessed by ICP-MS analysis. Control groups were instead injected with distilled water following the same procedures. Figure presents mean ± SEM calculated from *n* = 3 samples in control groups (dH2O) and *n* = 6 in treatment groups (1 mg/kg B.W. CoCl2). \* significantly different between control group and treatment group at a given time-point as assessed by two sample *t*-test (*p* < 0.05).

Several tissues in Figure 2 presented increased cobalt accumulation from 7- to 28-days, indicating possible cobalt time-dependent accumulation in the heart, liver, kidney, spleen, pref. cortex, and blood. However, many of the control samples in the 7-day treatment indicated high levels of cobalt (heart, liver, kidney, lung, spleen, testes, pref. cortex, and cerebellum). This augmentation is likely an artefact of ICP-MS measurements. For values closer to the Co detection limit, it could be that cobalt content is pushed to higher values.

Figure 3 shows the cobalt content detected by ICP-MS in all organs after 28 days of cobalt dose-treatment: 0.1, 0.5, and 1 mg/kg B.W. CoCl2. This metal content analysis revealed an incremental accumulation of cobalt concentrations in tissues with increased doses. Thus, there is a dose-response accumulation of cobalt, which was significant after 0.5 mg/kg B.W. CoCl2 in most tissues. Kidney, liver, and heart in that order accumulated most cobalt in line with the time-response experiment trend (Figure 2). The pref. cortex and hippocampus also accumulated significant levels of cobalt (*p* < 0.01). The same issue was observed with regard to the offset in the control samples, which were closer to the detection limit of the ICP-MS due to their low cobalt content. This effect became obvious in the case of the cerebellum control group.

**Figure 3.** Organ cobalt content (μg/g, tissue; μg/L, blood) obtained by ICP-MS after tissue and blood collection. SD male rats were treated with dH2O (control group) or different doses of CoCl2: 0.1, 0.5, and 1 mg/kg B.W. Animals were dosed daily with i.p. injections for 28 days. Each group presents mean ± SEM from *n* = 4 rats, and \* significant differences in control and treatment means as tested by one-way ANOVA (*p* < 0.05).

### *3.2. The Transcriptional Response to the Cobalt Doses Selected Is Non-Proportional*

The cobalt tissue content measured through ICP-MS determined that the pref. cortex and hippocampus of 0.5 and 1 mg/kg BW CoCl2-treated groups had significantly greater accumulated cobalt compared to their control groups (Figure 3). However, to evaluate whether the incremental dose cobalt treatment resulted in a progressive transcriptomic response, the number of genes was plotted for all brain areas and the three cobalt treatments: 0.1, 0.5, and 1 mg/kg B.W. CoCl2 in Figure 4 (threshold = |fold change| > 2). Although the number of DEGs progressively increased in the hippocampus, there was no clear dose-response in terms of the number of DEGs in the pref. cortex and cerebellum.

**Figure 4.** Number of upregulated (red) and downregulated (blue) DEGs (cutoff |fold change| > 2 only) in the pref. cortex, cerebellum, and hippocampus according to cobalt dose treatment: 0.1, 0.5, and 1 mg/kg B.W. CoCl2. Animals were dosed i.p. daily for 28 days with those doses or dH2O. Data were extracted from RNA-Seq experiments in which *n* = 4 samples were pooled to obtain *n'* = 1, except in the case of the hippocampus treatment group 0.5 mg/kg B.W. CoCl2, where *n'* = *n* = 3 as well as for 1 mg/kg B.W. CoCl2 where *n'* = *n* = 1.

Figure 5 shows the Venn diagram of DEGs found in the pref. cortex, cerebellum, and hippocampus. The diagram intersections display common DEGs between groups, with the pref. cortex and hippocampus demonstrating a higher number of common genes than the cerebellum for every dose. Although the number of overlapping genes between brain areas indicate that the hippocampus, cerebellum, and pref. cortex are all part of the same tissue (i.e., the brain), the different DEGs also point towards regionalised areas with specific functions. It can also be appreciated that there were few differences in the number of common genes between CoCl2 dose regimes. Moreover, it was not possible to identify sets of DEGs that followed a dose-response fold change after hierarchical clustering of the common genes (Figure S3). Thus, we can conclude that the number and range of the doses used does not prompt a dose-response. This does not mean that the transcriptional response elicited is not relevant.

### *3.3. Global Transcriptional Response in the Pref. Cortex and Hippocampus*

To further explore the effects on gene expression according to the dosage, this study focused on common DEGs for tissues with significant metal content accumulation. According to the results from the ICP-MS analyses, these tissues are the pref. cortex and hippocampus of rats dosed with 0.5 and 1 mg/kg B.W. CoCl2 (Figure 3). The result of this hierarchical clustering is displayed in Figure 6. The gene enrichment analysis of these genes generated with Cytoscape is displayed in Table 1. Several GO terms of importance involved in immunity and hormone activity could be observed. In addition, the protein–protein interaction (PPI) of DEGs–protein products was created to observe the possible links between the overlapped DEGs (Figure 7). The immune axis centred around interleukin-6

(IL6) was clearly separated from other clusters related to growth factors and hormone activity as well as some UDP-glucuronosyltransferases, specifically UGT enzymes that are involved in glucuronidation. Another cluster of interest reflected on the PPI is the glycosylphosphatidylinositol (GPI) anchor biosynthesis.

**Figure 5.** Venn diagrams showing the number of overlapping DEGs between the pref. cortex, cerebellum, and hippocampus at the different cobalt treatment doses: 0.1, 0.5, and 1 mg/kg B.W. CoCl2. Rats were treated by daily i.p. injection for 28 days. DEGs were obtained through RNA-Seq by comparing the brain parts' mRNA abundance of the treatment groups against the controls (dH2O-treated).

**Figure 6.** Hierarchical clustering of DEGs from brain tissues with significant accumulation of cobalt: the pref. cortex and hippocampus from rats treated with 0.5 and 1 mg/kg B.W. CoCl2. DEGs were obtained from RNA-Seq comparing the RNA isolated from those tissues with those of controls treated with dH2O. Condition applied is for fold change to be over 2. Upregulated genes are shown in red while downregulated are displayed in blue. Hierarchical clustering and resulting dendrogram were generated with Euclidian distance. Samples analysed through RNA-Seq were pooled (*n'* = 1) from *n* = 4 pref. cortex samples, *n* = 3 in hippocampus from 0.5 mg/kg B.W. CoCl2 treatment group, and *n* = 1 from 1 mg/kg B.W. CoCl2 treatment group.

**Table 1.** Enriched GO terms obtained from DEGs of the pref. cortex and hippocampus in response to cobalt treatment with 0.5 and 1 mg/kg B.W. CoCl2 compared to control animals (dH2O). Rats were dosed for 28 days with i.p. injections. GO terms annotated were significantly enriched with *p* < 0.05. GO terms enriched belong to the Molecular Function (MF; 8 April 2016), Biological Process (BP; 8 April 2016), Cellular Component (CC; 8 April 2016), and KEGG (14 June 2016) GO databases.


To better observe genes possibly involved in cobalt toxic mechanisms, the DEGs expressed over 2-fold change from the pref. cortex and hippocampus were plotted, as shown in Figures 8 and 9, respectively. For display purposes only, DEGs with a significance threshold <0.05 obtained from the Poisson distribution and provided in the original RNA-Seq data are shown. These are displayed as arranged by hierarchical clustering analyses of the DEGs (dendrogram not shown). Figure 8 displays the common genes of the pref. cortex across the three doses, and presents several genes whose proteins have a role in inflammation and immunity such as *Crp*, *Tnf*, and *Cxcl13*. Figure 9 shows significant DEGs of the hippocampus over 2-fold change. Surprisingly, there are several markers attributed to the choroid plexus such as *Clic6*, *Ttr*, *Kl*, *Col8a1*, and others [38–40].

**Figure 7.** Protein–protein interaction (PPI) network obtained from the STRING web tool by analysing DEGs as their protein products. DEGs were obtained from RNA-Seq analyses of the pref. cortex and hippocampus from rats treated with 0.5 and 1 mg/kg B.W. CoCl2 against controls treated with dH2O for 28 days of i.p. injections. The following terms/keywords have been highlighted: cellular response to interleukin-6 (blue), regulation of hormone levels (purple), chemokine receptors bind chemokines (yellow), blood coagulation (red), UDP-glucuronosyltransferase activity (pink), growth factor activity (green), and post-translational modification: synthesis of GPI-anchored protein (cyan). The thickness of links between nodes represent the confidence in the interaction, only nodes connected with high confidence (0.7) are displayed.

**Figure 8.** DEGs obtained from the comparison of pref. cortex from rats dosed with 0.1, 0.5, and 1 mg/kg B.W. CoCl2 against the control group (dH2O). Animals were treated for 28 days with daily i.p. injections. DEGs displayed were obtained from the RNA-Seq analysis of pooled samples (*n'* = 1 from *n* = 4 samples per group). Fold-change gene expression is indicated by colour, as described by the bar in the right side, upregulated genes are displayed in red while downregulated are blue. Genes are displayed as determined by the hierarchical clustering of DEGs over 2-fold-change (*p* < 0.05 from RNA-Seq Poisson distribution), dendrogram not shown (Euclidian distance).

A number of genes were evaluated via RT-qPCR according to their high expression level in RNA-Seq data as well as their function. The genes selected were *Tnf*, *Spata18*, *Ttr,* and *Akap14* in the pref. cortex and *Kl* in the hippocampus (see Figures S4 and S5 in the Supplementary Materials). In general, the RT-qPCR gene expression results agreed with the RNA-Seq data, and the fold expression of evaluated DEGs from RNA-Seq and RT-qPCR in this study did correlate. This is consistent with previous high-throughput comparisons between the two technologies [41] as well as research with pooled samples [42]. Thus, it was considered that the RT-qPCR results validate the RNA-Seq results.

**Figure 9.** DEGs expressed in hippocampus of rats treated via i.p. with daily injections of 0.1, 0.5, and 1 mg/kg B.W. CoCl2 or dH2O (control groups) for 28 days. Pooled samples (*n'* = 1 from *n* = 4 samples in group 0.1 mg/kg B.W. CoCl2, *n* = 3 from 0.5 mg/kg B.W. CoCl2, and *n* = 1 from 1 mg/kg B.W. CoCl2 group) were analysed through RNA-Seq and data are presented as the result of hierarchical clustering. DEGs in the graph are only those with fold-change >2 and *p* < 0.05 from RNA-Seq Poisson distribution, dendrogram is not shown. Colour bar presents fold-change: upregulated genes in red and downregulated genes in blue.

### **4. Discussion**

*4.1. Metal Distribution Pattern Is Consistent with Previous Research and Cobalt Accumulates Significantly in the Brain*

Figure 2 shows that there is a time-dependent cobalt accumulation trend in most tissues. This time-dependent trend might indicate that the longer cobalt levels remain elevated in blood, with a higher deposition of cobalt in nearly all organs tested, particularly in the heart, liver, and kidney. The accumulation of cobalt in most organs also appears to

be proportional to the dose used, as displayed in Figure 3. In addition, cobalt levels were significantly increased in the pref. cortex and hippocampus of rats dosed with 0.5 and 1 mg/kg B.W. CoCl2 at 28 days.

The liver, kidney, and heart accumulated more metal than other organs (Figures 2 and 3), which correlates with cobalt organ concentration in humans following cobalt radioisotope distribution [13]. This cobalt distribution follows the same pattern as previous experiments conducted by this research group [43,44] and other teams investigating cobaltism [45]. Most cobalt is excreted through urine [13,46], and given that both kidney and liver are involved in toxin and waste detoxification, it is not surprising that cobalt is predominantly concentrated in these tissues.

With regard to the cobalt in blood after the 28-day treatment with 1 mg/kg B.W., the level was 27.15 ± 2.70 μg/L in the time-response experiment while it was 38.24 ± 2.14 μg/L in the dose-response experiment. Given that the blood cobalt levels were under 100 μg/L, only subtle or moderate neurotoxicity was expected [1].

The cobalt content of two hearts from MoM patients severely affected by systemic cobalt toxicity has been reported in the literature as 4.75 [47] and 8.32 μg/g [48] (reference values for heart cobalt content: 0.06 μg/g) [49]. In addition, other post-mortem analyses on patients with metal-on-polyethylene hip prostheses also revealed significantly elevated cobalt content averaging 0.12 μg/g (range: 0.006–6.299 μg/g) [49]. Table 2 shows a compact presentation of these figures and results for comparison. The higher values of 0.34 ± 0.03 μg/g and 0.44 ± 0.06 μg/g in the rats' hearts at 28 days with 1 mg/kg B.W. CoCl2 treatment were also close to the average content in asymptomatic metal-onpolyethylene (MoP) patients (0.12 μg/g) [49]. Hence, the gradual dosing model used by this study and in previous studies [43] is comparable to the long-term systemic exposure of cobalt through circulating blood in MoM patients.

**Table 2.** Cobalt concentrations of whole blood (WB), brain and heart tissues in the dose and timeresponse experiments and other studies that mimic gradual cobalt release through daily treatment [45], cobalt tissue analyses in the cardiac tissue, and serum of MoM patients with cobaltism [47,48] as well as post-mortem heart tissue from metal-on-polyethylene (MoP) patients [49]. Cobalt content values in unexposed human brain (<0.025 μg/g), heart (0.060 μg/g) and blood (<1 μg/L) were obtained from [36,49,50] in that order. The abbreviations are: i.p., intraperitoneal injections; WB, whole blood; avg., average; C, control; T, treatment; THA, total hip arthroplasty. \* significantly different control and treatment groups as assessed by one-way ANOVA with Dunnett's multiple comparison.


To our knowledge, there is no information on the cobalt concentrations of MoM patients in the brain, probably due to the difficulty of obtaining samples from patients. In this study, the pref. cortex, and hippocampus had significant cobalt accumulated at 28 days (Figures 2 and 3). A study by Apostoli et al. with rabbits dosed with cobalt for

18 days intravenously reported elevated brain levels, 0.2 ± 0.2 μg/g, from the average control levels, 0.06 ± 0.04 μg/g dry weight, and cobalt whole blood, 420.9 ± 154.5 μg/L (see Table 2 for reference [45]). The histology in several organs only reported damage to the eyes and the auditory systems. The model used for this study had much lower cobalt blood concentration and animals remained in good health, while Apostoli et al. described balance disturbance in rabbits due to vestibular damage. Given the literature and observations of this study, cobalt appeared to induce only subtle or moderate neurotoxicity in this study even when cobalt had significantly accumulated in the rats' brains.

### *4.2. The Low Range Cobalt Dosage Used Does Not Lead to a Dose-Response*

No dose-response was demonstrated either in the number or in the average fold change of DEGs elicited by cobalt treatment (Figure 4 and Figure S3). Previous studies suggest that the fraction of ionic cobalt remains constant throughout a wide range of cobalt concentrations in the blood due to its albumin binding capacity [13]. If cobalt is being sequestered by albumin, it is likely that the response to any administered cobalt treatment will be dampened. However, it is also possible for cobalt toxic effects to follow a dose response, but the few concentrations used here did not cover such a range of toxicity.

### *4.3. Overall Transcriptional Effects of Cobalt and the Choroid Plexus as a Target of Cobalt Toxicity*

This study found some GO terms of interest that could be associated with cobalt toxicity in Table 1. This is the case of 'steroid hormone biosynthesis', which is comprised of a few genes of the Cytochrome P450 (*Cyp* prefix) family as well as by a couple UDPglucuronosyltransferases (*Ugt* prefix) and a sulfotransferase (*Sult* prefix). The protein products of these gene families form part of the drug metabolism and detoxification pathways. CYPs are also involved in the biosynthesis of serotonin, dopamine [51], and steroid hormones [52]. The findings suggest that CYPs could be a target of cobalt since they normally bind to the heme metal substrate as some metal ions have been observed to inactivate members of the CYP family [53]. Nevertheless, there are other large GO terms linked with hormone homeostasis such as 'regulation of hormone levels', 'steroid binding', and 'hormone activity' in Table 1. Some of these are not directly related to steroid hormones e.g., essential thyroid related genes (*Tshb*), whose deletion leads to hypothyroidism, and other GO terms such as 'response to retinoic acid'. Thus, nuclear receptors could possibly be regulated by cobalt. Nuclear receptors may bind steroids, retinoic acid, or thyroid hormones, and this binding depends on the nuclear receptor zinc finger domain. The CYP family synthesises some of the ligands that the nuclear receptors bind to, and further experiments could be performed to ascertain whether cobalt binds to CYPs or to nuclear receptors [54].

The PPI analysis (Figure 7) displayed a regulation of hormone levels consistently with the GO enrichment analysis as well as a network of drug-metabolising enzymes consisting of a few *Cyp*, *Ugt*, and *Sult* genes. The immune and haematopoietic axis centred around *Il6* with important chemokine presence can also be observed. This immune response was also reflected in the GO enrichment from Table 1. Some of the immune-related GO terms are 'response to interleukin-6- , 'T-helper 17 cell lineage commitment', 'response to interleukin-1- , 'cell chemotaxis', and 'cytokine receptor binding', which suggest immune cell differentiation, activity, and migration. Other factors related to blood coagulation such as *Hrg*, *Serpind1*, and *Fga* are displayed in Table 1. Activation of the immune system and dysregulation of haematopoietic transcriptional programmes could lead to several autoimmune and blood disease syndromes. Cobalt was a historical treatment of anaemia, and polycythaemia and skin rashes have been documented as a sporadic result of cobalt treatment [13]. Finally, the PPI also reported the presence of 'GPI-anchored protein' related transcripts. This is a post-transcriptional modification that mainly occurs in the endoplasmic reticulum, where most glycosylphosphatidylinositol (GPI) synthesis proteins function [55]. GPI-anchored protein synthesis depends on phospholipids, while the synthesis of steroid hormones is determined by cholesterol, hence, it is suggested here that cobalt modulates lipid metabolism. It is possible that cobalt could modulate lipid metabolism directly as cobalt has been seen to affect the rigidity of lipid membranes such as liposomes [56]. Lipid droplets were present in the spleen of a patient with a CoCr prosthesis [57]. Moreover, intracytoplasmic lipid and lipofuscin accumulation were found in a recent heart biopsy of a patient with arthroprosthetic cobaltism [48].

This study also found genetic markers in the pref. cortex (Figure 8) and hippocampus (Figure 9) almost exclusively attributable to the choroid plexus (e.g., *Clic6*, *Klotho* (*Kl*), transthyretin (*Ttr)*, *Veph1*, some cilia markers, and *Scl* transporters). The molecular characterisation of the choroid plexus has only been achieved recently [38], and unfortunately the GO ontologies have not been adequately updated to report its presence. The choroid plexus is anatomically attached to the hippocampus and its joint dissection can go unnoticed when doing a fast isolation, as reported by specialists in the choroid plexus [39,58]. In fact, several studies investigating the effect of drugs or other interventions in the hippocampus have knowingly or unwittingly reported choroid plexus markers [58–60]. There are also markers of the choroid plexus such as transthyretin (*Ttr*) in the pref. cortex, and it is possible that part of the choroid plexus has also been included in brain samples other than the hippocampus, since the choroid plexus is distributed through all brain ventricles [58]. Different studies have revealed that heavy metals preferentially accumulate early on in the choroid plexus, which appears to retain them, thus protecting the brain [61–66]. Harrison-Brown et al. discovered that the penetration of cobalt in the cerebrospinal fluid (CSF) of MoM patients was limited to 15% of the cobalt in plasma [67]. They also found a nonlinear trend with a ceiling effect in the CSF cobalt accumulation in relation to Co plasma levels in blood. Thus, the choroid plexus could function as an absorptive barrier, and early cobalt accumulation and damage in the brain might occur in the choroid plexus.

The choroid plexus is also a place for steroid hormone biosynthesis [68] and it hosts metabolising enzymes to deal with and metabolise xenobiotics [69]. Many immune cells are resident in the choroid plexus, which works as the site for immune trafficking with the brain [70]. B and T lymphocytes can infiltrate the choroid plexus and affect its function under certain challenges, thus leading to inflammation [71–73]. In particular, the 'cell chemotaxis' GO term includes *Cxcl13*, a chemokine that recruits B lymphocytes, and it has been involved in lymphoid infiltration in the choroid plexus in a mouse model of neuropsychiatric lupus [73]. Moreover, very recently, modulation of *Otx2* expression in the choroid plexus has been seen to regulate anxiogenic behaviour in mice [74], and in two studies, the choroid plexus transcriptome quickly responded to stress tests in mice [39,58]. Given that the choroid plexus has been implicated in depression disorders [75] and that some patients with elevated cobalt levels in their blood showed signs of neuropsychiatric symptoms such as depression [8,10], one might speculate that cobalt toxicity in the choroid plexus could impair its function, and contribute towards mood dysregulation.

### **5. Conclusions**

In summary, although the rat pref. cortex and hippocampus accumulated lower amounts of cobalt than other tissues, these accumulations were still significant at similar elevated circulating cobalt levels to those found in some patients with MoM implants. We found that the common transcriptional response to cobalt in the brain areas analysed involved hormone and drug-metabolising activity, in addition to also describing a powerful immune response, perhaps mediated by inteleukin-6 (IL-6). An underlying dysfunction in lipid metabolism is also likely. We suggest that these mechanisms could be instigated as a consequence of cobalt ion binding and substitution of native metal ions of CYPs or nuclear receptors. In the future, researchers should consider evaluating the markers of inflammation and lymphoid cell activation as well as the steroidogenic activity in the choroid plexus in response to cobalt. Thus, we have generated a mechanistic hypothesis for cobalt neurotoxicity that could be explored further and have relevant implications for patients with MoM implants who develop neurological health issues.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/ 10.3390/toxics10020059/s1, Table S1. Quality check of RNA samples from the pref. cortex tissue obtained from the dose-response in vivo experiments, Table S2. Quality check of RNA samples from the cerebellum tissues obtained from the in vivo dose-response experiments, Table S3. Quality check of RNA samples from the hippocampus tissues obtained from the in vivo dose-response experiments, Table S4. MIQE checklist from the MIQE guidelines for reproducibility and assessment of experimental RT-qPCR conditions, Table S5. Criteria for the design of the primers selected through NCBI Primer-BLAST tool, Table S6. Primer sequences of targeted genes designed for the in vivo dose-response experiment, Table S7. Fast PCR thermal cycling steps based on PowerUpTM SYBRTM Green Master Mix instructions for StepOnePlus Real-Time PCR system, Table S8: Primer sequences of control genes, Figure S1. RefFinder ranking of three reference genes Ct values in pref. cortex samples from the in vivo dose-response experiment, Figure S2. RefFinder ranking of three reference genes Ct values in hippocampus samples from the in vivo dose-response experiment, Figure S3. Hierarchical clustering of DEGs over 2-fold-change from RNA-Seq data obtained from the comparison of the pref. cortex, cerebellum, and hippocampus of rats treated with three concentrations of cobalt against those of controls treated with dH2O, Figure S4: Fold change mRNA gene expression levels obtained from RNA-Seq and RT-qPCR, Figure S5: Quantification of *Tnf*, *Spata18*, *Ttr*, and *Akap14* delta (CT) values (CT target gene–CT internal control) in the pref. cortex and *Kl* in the hippocampus through RT-qPCR.

**Author Contributions:** Conceptualization, S.G.-A., R.J.T. and M.H.G.; Methodology, S.G.-A., R.J.T. and M.H.G.; Software, S.G.-A.; Validation, S.G.-A.; Formal analysis, S.G.-A.; Investigation, S.G.-A.; Resources, R.J.T. and M.H.G.; Data curation, S.G.-A.; Writing—original draft preparation, S.G.-A.; Writing—review and editing, S.G.-A., R.J.T. and M.H.G.; Visualization, S.G.-A.; Supervision, R.J.T. and M.H.G.; Project administration, M.H.G.; Funding acquisition, M.H.G. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by an Engineering the Future (ETF) studentship awarded by the University of Strathclyde (internal funding, Faculty of Engineering, University of Strathclyde).

**Institutional Review Board Statement:** All animal procedures were conducted under the UK Home Office project licenses, 60/4341 and PDE5626B67. Ethic Committee Name: UK Home Office and AWERB. Approval Code: 60/4341 and PDE5626B67 (UK Home Office licence numbers). Approval Date: 25th April 2017.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** All data underpinning this publication are openly available from the University of Strathclyde KnowledgeBase at https://doi.org/10.15129/cbbe122f-4ee1-4bfb-9ec7-4 6c664bf10f2 (accessed on 29 November 2021).

**Acknowledgments:** Katie Henderson, Sarunya Laovitthayanggoon, Laura Beattie, and Ibrahim Alanazi provided invaluable technical and human support during the experiments. We are also grateful for the assistance and technical support of the Biological Procedures Unit (BPU) from the University of Strathclyde.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

### **References**

