Identifying Adipogenic Chemicals: Disparate effects in 3T3-L1, OP9 and primary mesenchymal multipotent cell models

Abstract

3T3-L1 pre-adipocytes are used commonly to identify new adipogens, but this cell line has been shown to produce variable results. Here, potential adipogenic chemicals (identified in the ToxCast dataset using the Toxicological Priority Index) were tested for their ability to induce adipocyte differentiation in 3T3-L1 cells, OP9 cells and primary mouse bone marrow multipotent stromal cells (BM-MSC). Ten of the 36 potential adipogens stimulated lipid accumulation in at least one model (novel: fenthion, quinoxyfen, prallethrin, allethrin, pyrimethanil, tebuconzaole, 2,4,6-tris(tert-butyl)phenol; known: fentin, pioglitazone, 3,3’,5,5’-tetrabromobisphenol A). Only prallethrin and pioglitazone enhanced lipid accumulation in all models. OP9 cells were significantly more sensitive to chemicals known to activate PPARγ through RXR than the other models. Coordinate effects on adipocyte and osteoblast differentiation were investigated further in BM-MSCs. Lipid accumulation was correlated with the ability to stimulate expression of the PPARγ target gene, Plin1. Induction of lipid accumulation also was associated with reduction in alkaline phosphatase activity. Allethrin, prallethrin, and quinoxyfen strongly suppressed osteogenic gene expression. BM-MSCs were useful in coordinately investigating pro-adipogenic and anti-osteogenic effects. Overall, the results show that additional models should be used in conjunction with 3T3-L1 cells to identify a broader spectrum of adipogens and their coordinate effects on osteogenesis.

1. Introduction

Adipose, along with the brain, liver and pancreas, is a critical contributor to the regulation of metabolic homeostasis. Adipocytes differentiate from multipotent stromal cells (MSC) capable of forming adipocytes, chondrocytes and osteocytes. Adipogenesis can be induced by exposure to insulin, inducers of cAMP signaling and glucocorticoids, which upregulate peroxisome proliferator activated receptor γ (PPARγ)¹ and induce the generation of an endogenous ligand, or by exposure to a potent PPARγ ligand alone (as reviewed in a). PPARγ forms a heterodimeric complex with retinoid X receptors (RXR) and binds to PPAR-response elements (5’-AACTAGGNCA A AGGTCA-3’). Ligand binding initiates a conformational change that results in the dissociation of corepressors and the association of coactivators, allowing ligand-induced transactivation (as reviewed in (Lefterova and Lazar, 2009)). PPARγ is a permissive heterodimer partner of RXR, thus RXR ligands also can stimulate PPARγ-dependent adipocyte differentiation (Schulman et al., 1998). However, permissiveness can be limited, with efficacy dependent upon the heterodimerization partner, the cell type and the gene target (Szeles et al., 2010).

The hunt for metabolism disrupting chemicals (MDCs), so named in the Parma consensus statement on metabolic disruptors (Heindel et al., 2017), has largely focused on identifying ligands of peroxisome proliferator activated receptor γ (PPARγ) (Tontonoz et al., 1994). PPARγ is a ligand activated, nuclear receptor and essential regulator of adipocyte formation and function (Tontonoz et al., 1994), as well as metabolic homeostasis, as all PPARγ haploinsufficient and knockout models present with lack of adipocyte formation and metabolic disruption (He et al., 2003; Zhang et al., 2004; Jiang et al., 2014; Gumbilai et al., 2016; O’Donnell et al., 2016; Gilardi et al., 2019)). Further, excessive activation of PPARγ is required for a high fat diet to induce adipocyte hypertrophy and insulin resistance (Kubota et al., 1999)). A growing number of environmental PPARγ ligands, which stimulate adipocyte differentiation, have been identified (e.g. organotins, phthalates, parabens, tetrabromobisphenol A, organophosphate esters, and pesticides) (Kanayama et al., 2005; Grun et al., 2006; Kirchner et al., 2010; Riu et al., 2011; Li et al., 2012; Hu et al., 2013; Pillai et al., 2014; Heindel et al., 2017). Furthermore, tributyltin, di(2-ethylhexyl)phthalate, triflumizole and Firemaster® 550, a flame retardant mixture containing organophosphate esters, all have been shown to increase fat mass in vivo (Feige et al., 2010; Li et al., 2012; Patisaul et al., 2013).

Human treatment with therapeutic PPARγ ligands (e.g. thiazolidinediones) is associated with both an increased fat mass and a risk of bone fracture (Miyazaki et al., 2001; Billington et al., 2015). MSC differentiation is controlled by the balance of PPARγ and Runt-related transcription factor 2 (Runx2, master regulator of bone formation (Ducy et al., 1997)) transcriptional activation and co-regulated by signaling through the Wnt/β-catenin pathway, which modulates the expression and function of both PPARγ and Runx2 (Bennett et al., 2005; Jeon et al., 2003; Kang et al., 2007; Moldes et al., 2003). Runx2 expression and activity increase upon commitment to the osteoblast lineage, leading to the activation of a second essential transcription factor, osterix (Liu and Lee, 2013). An increase in PPARγ activation suppresses osteogenesis and drives differentiation toward adipogenesis (Kirchner et al., 2010; Lecka-Czernik et al., 1999). Conversely, decreasing expression of PPARγ (e.g. by molecular knockdown) enhances expression of Runx2 and osterix, increases bone mass and decreases adipogenesis (Akune et al., 2004).

A common approach to identifying adipogenic chemicals is testing in the 3T3-L1 preadipocyte model. 3T3-L1 preadipocytes have been used as a model to study the process of adipocyte differentiation since the 1970s (Green and Kehinde, 1975) and are a model system that has been used in close to 10,000 publications. A Pubmed search of the terms “3T3-L1 environmental adipogenic” results in 92 hits. A screening system for identifying obesogenic chemicals using 3T3-L1 cells also has been proposed (Pereira-Fernandes et al., 2013). However, the efficacy of activation of PPARγ is dependent upon cell type (Sewter et al., 2002; Feige et al., 2007; Lefterova et al., 2010; Sarusi Portuguez et al., 2017) and likely results from cell-specific differences in PPARγ expression, chromatin structure and coactivator availability (Spiegelman and Heinrich, 2004; Sarusi Portuguez et al., 2017). The fine tuning of PPARγ activity is evident from studies of differential activation of PPARγ by a partial agonist at distinct stages of adipocyte differentiation in 3T3-L1 cells (Fujimura et al., 2006). Furthermore, a recent study has shown that the cell model used (3T3-L1 cells vs. OP9 cells) can significantly impact the detection of chemicals that induce adipogenesis (Kassotis et al., 2017).

As a first step in determining which environmental PPARγ ligands may be contributing to the obesity epidemic, we need to know the full spectrum of chemicals in the environment that can activate the PPARγ pathway. To identify these ligands, we need an approach that is biologically relevant and robust to the multiple ligand- and cell-dependent effects on PPARγ. Recent reviews have outlined the different models that are used to identify and study adipogens (Ruiz-Ojeda et al., 2016; Chamorro-Garcia and Blumberg, 2019). Here, we directly compare the response of 3T3-L1 cells (Green and Kehinde, 1975), OP9 cells (Wolins et al., 2006) and primary bone marrow multipotent cells (BM-MSCs) to known adipogenic chemicals, as well as chemicals predicted to be adipogenic by the ToxPi approach. These models represent different stages of commitment to the adipocyte lineage, with BM-MSCs maintaining the ability to differentiate into either adipocytes or osteocytes (Watt and Schlezinger, 2015), 3T3-L1 cells being a pre-adipocyte model (Lane and Tang, 2005) and OP9 cells being a late pre-adipocyte model (Wolins et al., 2006). Further, we investigated if pro-adipogenic activity was correlated with anti-osteogenic activity.

2. Materials and Methods

2.1. Materials

Rosiglitazone (Rosi) was from Cayman Chemical (Ann Arbor, MI). Dimethyl sulfoxide (DMSO) was from American Bioanalytical (Natick, MA). Insulin, Nile Red, p-nitrophenyl phosphate (pNPP) reagent, LG100268 (LG268), LG100754, tributyltin chloride (TBT), and triphenyl phosphate (TPhP) were from Sigma-Aldrich (St. Louis, MO). Test chemicals (Table 1) also were from Sigma-Aldrich, provided via the National Toxicology Program or ordered directly. LG100754 was from (Bio-techne Corp., Minneapolis, MN). All other reagents were from Thermo Fisher Scientific (Suwanee, GA) unless noted.

Table 1.

Test chemical characteristics and adipogenic activity (lipid accumulation).

Chemical Name	Abbr	Concs Tested	Use	Positive Analyses in ToxCast	Max Tolerated Conc	% Max Lipid Accuma	Max Tolerated Conc	% Max Lipid Accuma	Max Tolerated Conc	% Max Lipid Accuma
(CAS #)					3T3 L1		OP9		BM-MSC
Negative Controls
Acetamiprid 135410-20-7	Aceta	10 nM − 40 μM	Insecticide		40 μM	—	40 μM	—	20 μM	—
Asulam 3337-71-1	Asula	10 nM − 40 μM	Herbicide		40 μM	—	40 μM	—	10 μM	—
Flumetsulam 98967-40-9	Flume	10 nM − 40 μM	Herbicide		40 μM	—	40 μM	—	20 μM	—
Maleic hydrazide 123-33-1	MalHy	10 nM − 40 μM	Herbicide		40 μM	—	40 μM	—	20 μM	—
Methamidophos 10265-92-6	Metha	10 nM − 40 μM	Insecticide		40 μM	—	40 μM	—	1 μM	—
Pymetrozine 123312-89-0	Pymet	10nM −40 μn	Insecticide		40 μM	—	40 μM	—	20 μM	—
Methylene bis(thiocyanate) 6317-18-6	MBTC	10 nM − 20 μM	Microbiocide		2 μM	—	2 μM	—	20 μM	—
Test Chemicals
(Z,E)-Fenpyroximate 134098-61-6	Fenpy	10 nM − 20 μM	Acaricide	LXR, PPARγ	20 μM	—	20 μM	—	10 μM	—
Bentazon 25057-89-0	Benta	2 – 20 μM	Herbicide	CEBPβ, LXR, PPARγ	20 μM	—	20 μM	—	20 μM	—
Norflurazon
27314-13-2	Norfl	2 – 20 μM	Herbicide	SREBF1	20 μM	—	20 μM	—	20 μM	—
3,3’,5,5’-
Tetrabromobi sphenol A 79-94-7	TBBPA	2– 20 μM	Flameretardant	GR, PPARγ, SREBF1	20 μM	↑	20 μM	↑	20 μM	↑
Citric Acid 77-92-9	CitAc	2 – 20 μM	Food ingredient	PPARγ, RXR	20 μM	—	20 μM	—	20 μM	—
Methyl salicylate 119-36-8	MeSal	2 – 20 μM	Food ingredient	CEBPβ, LXR, PPARγ, RXR	20 μM	—	20 μM	—	5 μM	—
Cyazofamid 120116-88-3	Cyazo	10 nM − 20 μM	Fungicide	CEBPβ, LXR, PPARγ	20 μM	—	20 μM	—	20 μM	—
Fentin 76-87-9	FenHy	1 nM − 500 nM	Fungicide	PPARγ	50 nM	↑	50 nM	↑	12.5 nM	↑
Fluazinam 79622-59-6	Fluaz	10 nM − 20 μM	Fungicide	PPARγ	2 μM	↓	2 μM	—	1 μM	—
Fludioxonil 131341-86-1	Fludi	10 nM − 20 μM	Fungicide	CEBPβ, PPARγ, RXR	2 μM	—	2 μM	—	2 μM	—
Flusilazole 85509-19-9	Flusi	10 nM − 20 μM	Fungicide	GR, PPARγ, SREBF1	20 μM	—	20 μM	↓	10 μM	—
Imazalil 35554-44-0	Imaza	10 nM − 20 μM	Fungicide	RXR	20 μM	—	20 μM	↓	10 μM	↓
Pyrimethanil 53112-28-0	Pyrim	1 – 20 μM	Fungicide	CEBPβ, LXR, PPARγ, SREBF1	20 μM	—	20 μM	—	20 μM	↑
Quinoxyfen 124495-18-7	Quino	10 nM − 20 μM	Fungicide	PPARγ	4 μM	↑	4 μM	—	8 μM	↑
Tebuconazole 107534-96-3	Tebuc	2 – 20 μM	Fungicide	SREBF1	20 μM	↑	20 μM	—	20 μM	—
d-cis,trans-Allethrin 584-79-2	Allet	10 nM − 20 μM	Insecti cide	CEBPβ, PPARγ, SREBF1	10 μM	↑	10 μM	—	10 μM	—
Fenthion 55-38-9	Fenth	10 nM − 40 μM	Insecticide	LXR, PPARγ	40 μM	↑	40 μM	↑	20 μM	↑
Prallethrin 23031-36-9	Prall	10 nM − 20 μM	Insecticide	CEBPβ, PPARγ, SREBF1	20 μM	↑	20 μM	↑	20 μM	↑
Pyridaben 96489-71-3	Pyrid	1 nM − 100 nM	Insecticide	CEBPβ, PPARγ, SREBP1	20 nM	—	20 nM	—	20 nM	—
Rotenone 83-79-4	Roten	20 nM− 0.2 μM	Insecticide	CEBPβ, GR, LXR, PPARγ, SREBF1	20 nM	—	20 nM	—	20 nM	—
Tebufenpyrad 119168-77-3	Tebuf	2 nM−0.2 μM	Insecticide	CEBPβ, LXR, PPARγ, SREBF1	200 nM	—	200 nM	—	200 nM	—
All trans retinoic acid 302-79-4	ATRA	0.2 – 2 μM	Pharmaceutical	PPARγ, RXR	2 μM	↓	2 μM	↑	500 nM	↓
Corticosterone 50-22-6	Corti	0.2 − 2 μM	Pharmaceutical	GR	2 μM	↓	2 μM	—	500 nM	—
Dexamethazone 50-22-2	Dexam	20 nM − 0.2 μM	Pharmaceutical	GR	200 nM	↓	200 nM	—	20 nM	—
Ducosate sodium 577-11-7	DOSS	1 – 10 μM	Pharmaceutical	GR, PPARγ, RXR, SREBF1	10 μM	—	10 μM	—	10 μM	—
17α-Hydroxyprog esterone 68-96-2	17aHP	2 – 20 μM	Pharmaceutical	GR, SREBF1	20 μM	—	20 μM	—	10 μM	—
Pioglitazone 112529-15-4	Piogl	1 – 10 μM	Pharmaceutical	PPARγ, SREBF1	5 μM	↑	5 μM	↑	2.5 μM	↑
Simvastatin 79902-63-9	Simva	2 – 20 μM	Pharmaceutical	GR, SREBF1	1 μM	—	1 μM	—	100 nM	—
Triamcinolone 124-94-7	Triam	0.2 − μM	Pharmaceutical	GR	2 μM	—	2 μM	—	500 nM	—
Forclorfenuron 68157-60-8	FCF	10 nM − 40 μM	Pant growth regulator	CEBPβ, PPARγ	40 μM	—	40 μM	↓	10 μM	—
Bisphenol A 80-05-7	BPA	10 nM − 40 μM	Plastics component	GR, PPARγ	40 μM	—	40 μM	—	20 μM	—
Diallyl Phthalate 131-17-9	DiaPh	2 – 20 μM	Plastics component	CEBPβ, LXR, PPARγ, RXR	20 μM	—	20 μM	—	20 μM	—
N,N-Dimethylform amide 68-12-2	DMF	2 – 20 μM	Solvent	PPARγ, RXR	20 μM	—	20 μM	—	20 μM	—
Perfluoroocta ne sulfonic acid 1763-23-1	PFOS	10 nM − 40 μM	Surfactant	GR, PPARγ	40 μM	—	40 μM	—	20 μM	—
2,4,6-Tris(tertbutyl) Phenol 732-26-3	TTBP	2 – 20 μM	Chemical intermediate	PPARγ, RXR	20 μM	—	20 μM	↑	5 μM	—

2.2. Selection of potential adipogens

We identified a set of well-documented, synthetic and environmental PPARγ and RXR ligands (Rosi, LG268, LG100754, TBT, and TPhP) as known adipogens (Lehmann et al., 1995; Cesario et al., 2001; Grun et al., 2006; Pillai et al., 2014). For novel, potential adipogens (Table 1), we used chemicals identified by the Toxicological Priority Index (ToxPi) model for adipogenesis that was generated by expert-defined relevant biological processes and applied to the ToxCast Phase I and Phase II releases (Knudsen et al., 2011; Filer et al., 2014; Auerbach et al., 2016; Janesick et al., 2016; Judson et al., 2016). We also identified negative control chemicals from the ToxPi model (Table 1).

2.3. Cell culture

3T3 L1 (RRID:CVCL_0123) cells were originally derived from a Swiss mouse embryonic fibroblast line (Green and Kehinde, 1975). Cells were maintained in high-glucose DMEM (Corning, 10–013-CV) with 10% calf serum (Sigma), 100 U/ml penicillin, 100 μg/ml streptomycin, 0.25 μg/ml amphotericin B. All experiments were conducted with cells between passages 3 and 8. Cells were plated in 24 well plates at 50,000 cells per well and incubated for 4 days, at which time the cultures were confluent for 2 days. To induce adipogenesis, the medium was replaced with DMEM containing 10% fetal bovine serum (FBS, Sigma), 25 nM dexamethasone, 167 nM of 1 µg/ml human insulin, 0.5 mM IBMX, 100 U/ml penicillin, and 100 μg/ml streptomycin. Upon addition of induction medium, cells received no treatment (naïve) or were dosed with vehicle (Vh, DMSO, 0.2% final concentration), known adipogens (Rosi, 200 nM; LG268, 200 nM; LG100754, 200 nM; TBT, 100 nM; TPhP, 10 μM), or ToxPi chemicals as indicated in Table 1. Rosi also was used as a positive control. On days 3 and 5 of differentiation, medium was replaced with adipocyte maintenance medium (DMEM, 10% FBS, 167 nM human insulin, 100 U/ml penicillin, 100 μg/ml streptomycin), and the cultures were re-dosed. On Day 7 of differentiation, medium was replaced with adipocyte medium (DMEM, 10% FBS, 100 U/ml penicillin, 100 μg/ml streptomycin) and the cultures re-dosed. Cells were analyzed on day 10 of differentiation. Cells were dosed 5 times overall. Cytotoxicity was assessed by microscopic inspection. If cultures contained more than 10% rounded cells the experiment was repeated at a lower concentration (Table 1).

OP9 cells (RRID:CVCL_4398) are a bone marrow stromal cell line derived from newborn calvaria of the (C57BL/6XC3H)F2-op/op mouse (Nakano et al., 1994). Cells were maintained in αMEM (Gibco, 12–561-056) with 20% FBS, 26 mM sodium bicarbonate, 100 U/ml penicillin, 100 μg/ml streptomycin, 0.25 μg/ml amphotericin B. Cells were plated in 24 well plates at 50,000 cells were well in 500 µl medium and incubated for 4 days. Induction and maintenance of adipogenesis and treatment were as described for 3T3 L1 cells.

Primary bone marrow cultures were prepared from C57BL/6J mice (female, 12 weeks of age, RRID:IMSR_JAX:000664, Jackson Laboratories, Bar Harbor, ME). Studies were reviewed and approved by the Institutional Animal Care and Use Committee at Boston University. All animals were treated humanely and with regard for alleviation of suffering. Mice were housed 4 per cage, with a 12 hour light cycle. Water and food (2018 Teklad Global 18% Protein Rodent Diet, Irradiated; Harlan Laboratories, Indianapolis, IN) were provided ad libitum. Animals were euthanized for collection of bone marrow two days after arrival. Bone marrow was flushed from the femur, tibia and humerus bones, strained through a 70 μm cell strainer, diluted in MSC media (α-MEM containing 10% FBS and 100 U/ml penicillin, 100 μg/ml streptomycin, 0.25 μg/ml amphotericin B), and seeded at 6 million cells per well in 1 ml in 12 well plates. Half of the medium was replaced 4 days after plating, and the cultures incubated for 3 more days. To induce differentiation, the medium was replaced with MSC media containing 12.5 μg/ml ascorbate, 8 μM β-glycerol phosphate, 10 nM dexamethasone, and 500 ng/ml human insulin. Upon addition of induction medium, cultures received no treatment (naïve) or were treated with Vh (DMSO, 0.1% final concentration), known adipogens (Rosi, 200 nM; LG268, 200 nM; TBT, 100 nM; TPhP, 10 μM), or ToxPi chemicals as indicated in Table 1. Following treatment, the cells were cultured for 4–10 days. Medium was changed, and the cultures were redosed 3 times over a 7 day period (gene expression, lipid accumulation) or 5 times over a 12 day period (alkaline phosphatase activity, lipid accumulation).

2.4. Bone marrow cell phenotyping

Adherent cells were collected following 7 days of culture, prior to induction of differentiation. Cells were fixed, permeabilized and stained using BD Cytofix/Cytoperm according to the manufacturer’s instructions (BD Biosciences, San Jose, CA). Information on the antibodies is provided in Table S1. Cells were resuspended in cold FACS buffer and analyzed immediately on a BD™ LSRII Flow Cytometer at the BU Flow Cytometry Core Facility. Data were compensated and analyzed using FlowJo version 7 (Tree Star, Inc., Ashland, OR).

2.5. Adipocyte and bone phenotype assays

Lipid accumulation was quantified by Nile Red staining. 3T3 L1, OP9 and BM-MSCs were stained with an aqueous solution of Nile Red (1 μg/ml in phosphate buffered saline (PBS)) for 15 min, and fluorescence (excitation 485 (20 nm bandwidth), emission 530 nm (25 nm bandwidth) was measured using a Synergy2 multifunction plate reader (Biotek Inc., Winooski, VT) (Yanik et al., 2011). For cultures treated with positive controls, fluorescence in experimental wells was normalized by subtracting the fluorescence measured in untreated cultures and reported as relative fluorescence units (RFUs). For cultures treated with test chemicals, the naïve correct RFUs in the experimental wells were divided by the naïve corrected RFUs in 100 nM Rosi treated wells and reported as “% Rosi Control.”

Following Nile Red staining of 12 day BM-MSC cultures, cells were rinsed with PBS and fixed in paraformaldehyde (2% in PBS). To quantify alkaline phosphatase activity, cells were incubated in pNPP solution. After quenching with sodium hydroxide (final concentration: 0.75M), absorbance (405 nM) was measured using a Synergy2 plate reader. The absorbance in all experimental wells was normalized by dividing by the absorbance measured in wells that received osteogenic medium but were not treated and reported as “Fold Change.”

2.6. Reporter Assay

Cos-7 cells (in 96 well plates) were transiently transfected with vectors containing human PPARG1 (provided by V.K. Chatterjee, University of Cambridge, Cambridge, UK)(Gurnell et al., 2000) or the with pcDNA vector, PPRE x3-TK-luc (plasmid 1015; Addgene) (Kim et al., 1988), and CMV-eGFP using Lipofectamine 2000 (Invitrogen, Carlsbad, CA). Transfected cultures were incubated overnight. The medium was replaced, and cultures received no treatment (Naïve) or were treated with Vh (DMSO, 0.1%), Rosi or test chemicals (5 order of magnitude concentration span). After 24 hours, cells were lysed with Steady-Luc Firefly HTS Assay reagents (Cat # 30028, Biotium, Inc., Freemont, CA). Luminescence and fluorescence were determined using a Synergy2 plate reader. Luminescence of each well was normalized to GFP-fluorescence. The resulting values were normalized to that in wells treated with 100 nM Rosi and reported as “% Rosi Control.”

2.7. mRNA Expression

Following Nile Red staining, the staining solution was removed, and the cultures were washed two times with PBS. Then, total RNA was extracted and genomic DNA was removed using the RNeasy Plus Mini Kit (Qiagen, Valencia, CA). For RT-qPCR analyses, cDNA was prepared from total RNA using the GoScript™ Reverse Transcription System (Promega), with a 1:1 mixture of random and Oligo (dT)15 primers. All qPCR reactions were performed using the GoTaq® qPCR Master Mix System (Promega). Validated primers (18s ribosomal RNA (Rn18s): QT01036875, PPARγ1/2 (Pparg): QT00100296, fatty acid binding protein 4 (Fabp4): QT00091532, perilipin (Plin1): QT00150360, runt related transcription factor 2 (Runx2): QT00102193, Osterix (Osx): QT00293181, osteocalcin (Bglap): QT00259406O) were purchased from Qiagen. qPCR reactions were performed using a 7500 Fast Real-Time PCR System (Applied Biosystems, Carlsbad, CA). Relative gene expression was determined using the Pfaffl method (Pfaffl, 2001), using the threshold value for Rn18s for normalization. No significant differences were observed in the expression of Rn18s across the different treatments (data not shown). The Cq value from naïve, undifferentiated cultures was used as the reference point. Data were normalized to the expression in cultures treated with 100 nM Rosi and reported as “% Rosi Control.”

2.8. Statistics

Statistical analyses were performed with Prism 6 (GraphPad SoftwareInc., La Jolla, CA). Data are presented as means ± standard error (SE). Each n represents either an independent plating of 3T3 L1 or OP9 cells or an independent bone marrow preparation. Experimental data were compared to the Vh-treated cells included on the same plates to determine significance. All data from Vh-treated cells were aggregated in the figures. Gene expression data were log transformed prior to analysis. One-way ANOVAs with the Dunnett’s post hoc test, Student’s t test, linear regression and Pearson’s correlation analyses were performed where noted. All analyses were performed at α = 0.05.

3. Results

3.1. Assessment of chemical effects on pre-adipocyte models

The 3T3-L1 pre-adipocyte cell line is a classic model used in identifying adipogenic chemicals; however, significant variability in adipogenicity has been observed (Kassotis et al., 2017). Therefore, we tested the biological effects of the known and potential adipogens in both 3T3-L1 and OP9 cells. We began by testing a series of positive control chemicals known to directly activate PPARγ (Rosi, TPhP) or to indirectly activate PPARγ through RXR (LG268, TBT). Confluent 3T3-L1 cells and OP9 were stimulated to undergo adipogenesis and treated with Vh (DMSO, 0.1%) and positive control chemicals for 10 days. Hormone-induced adipogenesis (as indicated by lipid accumulation) in 3T3-L1 cells was significantly enhanced by Rosi and TPhP (Figure 1A). In OP9 cells, adipogenesis was significantly enhanced by Rosi, LG268, TBT and TPhP (Figure 1C).

Figure 1. Lipid accumulation induced by test compounds in 3T3 L1 and OP9 cell lines undergoing adipogenesis.

A standard hormonal protocol was used to induce adipogenesis (See Methods). At the initiation of differentiation, cells received no treatment (naïve) or were dosed with Vh (DMSO, 0.2% final concentration), positive controls (Rosi, 200 nM; LG268, 200 nM; TBT, 100 nM; TPhP, 10 μM), or test compounds as indicated in Table 1. Lipid accumulation was quantified by Nile Red staining on day 10. A) Positive controls in 3T3-L1. B) Test compounds in 3T3-L1. C) Positive controls in OP9. D) Test compounds in OP9. Hatched bars = Vh. Light gray bars = negative controls. Dark gray bars = experimental compounds. Data are presented as means ± SE from 3–5 independent biological replicates. Statistically different from Vh-treated (* p<0.05, ** p<0.01, ANOVA, Dunnett’s).

Second, we tested potential adipogenic chemicals (i.e., ToxPi chemicals) identified in the ToxCast dataset using the ToxPi method. Table 1 shows the abbreviations for all chemicals, the intended use of each chemical, the highest concentration tested and the highest non-toxic concentration, which was used in the differentiation analyses. In 3T3-L1 cells, the known PPARγ ligands pioglitazone (Piogl) and 3,3’,5,5’-tetrabromobisphenol A (TBBPA) significantly enhanced adipocyte differentiation, along with tebuconazole (Tebuc), fenthion (Fenth), quinoxyfen (Quino), prallethrin (Prall) and allethrin (Allet)(Figure 1B). Similarly, Piogl and TBBPA significantly enhanced adipocyte differentiation in OP9 cells (Figure 1D), along with retinoic acid (ATRA), Prall, Fenth, 2,4,6-tris(tert-butyl) phenol (TTBP) and fentin (FenHy). The less potent chemicals in the 3T3-L1 cells (Allet, Quino and Tebuc) did not significantly stimulate adipogenesis in OP9 cells (Figure 1D). The chemicals that antagonized adipogenesis were non-overlapping in the two cells lines. ATRA significantly reduced adipogenesis in 3T3 L1 cells, along with dexamethasone (Dexam), corticosterone (Corti) and fluazinam (Fluaz)(Figure 1B). In OP9 cells, imazalil (Imaza), forchlorfenuron (FCF) and flusilazole (Flusi) significantly reduced adipogenesis (Figure 1D).

To examine the association of responses to the known and potential adipogens in the two cell lines, we conducted a correlation analysis (Figure 2A). While the p value was significant (p = 0.0014), the correlation was weak (Pearson’s correction coefficient = 0.44). When known RXR ligands were removed from the analysis (ATRA, TBT, LG268, LG100754 and FenHy; Figure 2B), the p value increased (p < 0.0001), and the correlation became stronger (Pearson’s correction coefficient = 0.65). However, one outlier remained, TTBP.

Figure 2. Correlation of lipid accumulation induced in 3T3 L1 and OP9 cells.

Data are from Figure 1. Each data point represents the average lipid accumulation induced by a chemical (reported as “% Rosi Control) in each cell line. The green “X” is Vh. The red “X” is Rosi. Red circles indicate known RXR ligands. A) All compounds tested. B) Known RXR ligands removed. The dotted line represents the 90% prediction bands based on the linear regression.

Last, to confirm the identity of the adipogenic chemicals as PPARγ agonists, we tested them in a human PPARγ1-dependent reporter assay. As expected, allethrin, fenthion, fentin, pioglitazone, prallethrin, quinoxyfen, and TBBPA all significantly activated the transcriptional activity of PPARγ; however, only pioglitazone was a fully efficacious ligand (i.e., it induced 100% activity relative to Rosi; Figure 3). Neither tebuconazole nor TTBP activated PPARγ transcriptional activity (Figure 3).

Figure 3. PPARγ reporter activation by test compounds.

Cos-7 cells transfected with a human PPARγ1 expression construct and a PPRE-luciferase reporter plasmid were dosed with Vh (DMSO, 0.1% final concentration) or the indicated test compounds. Reporter activity was measured after 24 hours. Data were normalized as described in the Methods. Data are presented as means ± SE from 4 independent transfections. Statistically different from Vh-treated (* p<0.05, ** p<0.01, *** p<0.001, ANOVA, Dunnett’s).

3.2. Assessment of chemical effects on adipogenesis and osteogenesis in bone marrow cultures

A generally accepted paradigm is that adipogenesis and osteogenesis are largely mutually-exclusive (Lecka-Czernik et al., 1999). Thus, we examined concurrent effects on adipogenesis and osteogenesis using primary bone marrow cultures prepared from female C57BL/6J mice. First, we characterized the adherent cells generated following 7 days in culture, to ensure that cells capable of both osteogenesis and adipogenesis were generate, by phenotyping for MSC markers (Hegyi et al., 2010). After one week in culture, 49 ± 1% of the adherent cells were of non-hematopoietic origin (CD45^-, Ter119^-). Of the non-hematopoietic cells, 80 ± 1% expressed the MSC markers CD44 and Sca1. Only 5% of the MSCs expressed CD105, while >90% of the MSCs expressed αSMA (Figure S1). The strong expression of CD44 (a multifunctional cell surface protein involved in cell proliferation and differentiation) and Sca1 (a signal transduction protein necessary to maintain bone quality throughout life) is consistent with the ability of bone marrow- and spleen-derived MSCs to efficaciously undergo both adipogenesis and osteogenesis (Holmes and Stanford, 2007; Hegyi et al., 2010).

Since adipogenesis can be stimulated directly by a PPARγ ligand in the presence of FBS, dexamethasone and insulin, while osteogenesis requires ascorbate and β-glycerol phosphate, established bone marrow cultures were switched to medium containing ascorbate, β-glycerol phosphate, dexamethasone, and insulin prior to exposures. Cultures were treated with Vh (DMSO, 0.1%), positive control chemicals, or test chemicals as indicated in Table 1 for 7 days. Cultures treated with Rosi, TBT, and TPhP significantly accumulated lipid (Figure 4A). Seven of the 36 test chemicals (Piogl, TBBPA, Quino, Prall, FenHy, Fenth, pyrimethanil (Pyrim)) and none of the 7 negative controls significantly increased lipid accumulation (Figure 4B). Most chemicals were active in the 2–20 μM range. However, FenHy was more potent and induced significant lipid accumulation at 12.5 nM. TBBPA and Piogl, which are known PPARγ ligands, had the greatest efficacy in inducing lipid accumulation that was comparable to the control Rosi (Figure 4B). Notably, two chemicals (ATRA and Imaza) significantly decreased lipid accumulation.

Figure 4. Lipid accumulation induced by test compounds in primary mouse BM-MSCs undergoing osteogenesis.

After primary bone marrow cultures were established, differentiation was initiated with the addition of ascorbate, β-glycerol phosphate, insulin and dexamethasone. At the initiation of differentiation, cells received no treatment (naïve) or were dosed with Vh (DMSO, 0.2% final concentration), positive controls (Rosi, 200 nM; LG268, 200 nM; TBT, 100 nM; TPhP, 10 μM), or test compounds as indicated in Table 1. Lipid accumulation was quantified by Nile Red staining on day 7. A) Positive controls. B) Test compounds. Hatched bars = Vh. Light gray bars = negative controls. Dark gray bars = experimental compounds. Data are presented as means ± SE from 4 independent bone marrow preparations. Statistically different from Vh-treated (* p<0.05, ** p<0.01, ANOVA, Dunnett’s).

In order to confirm that lipid accumulation resulted from an adipogenic process, mRNA expression of adipogenic genes was determined following treatment with the 7 most potent, efficacious and broadly acting adipogens (Allet, Fenth, FenHy, Piogl, Prall, Quino, TBBPA), with the most efficacious inhibitor of adipogenesis (ATRA), and with a representative negative control chemical (pymetrozine (Pymet)). We quantified mRNA expression of total PPARγ, including isoforms 1 and 2, fatty acid binding protein 4 (Fabp4), a PPARγ target gene, and perilipin (Plin1), which regulates lypolysis and is expressed by differentiated adipocytes (Greenberg et al., 1991; Tontonoz et al., 1994). Pparg expression was significantly induced by Rosi (Figure S2A), but not by any of the test chemicals (Figure 5A). Pparg expression was decreased by the known adipogenic antagonist ATRA (Figure 5A). Fabp4 expression increased significantly with fentin, pioglitazone, prallethrin, and TBBPA (Figure 5B), and a similar pattern was seen with Plin1 expression (Figure 5C). Lipid accumulation was highly correlated with Plin1 expression (Figure 5D), suggesting that Nile Red fluorescence is an accurate method for assessing adipogenesis associated with PPARγ activation.

Figure 5. Adipogenic gene expression induction by active compounds.

Primary bone marrow cultures were established and differentiation was initiated as described in Figure 4. Cultures received no treatment (Naïve) or were treated with Vh (DMSO, 0.1% final concentration), Rosi (positive control, 100 nM) or test compounds as indicated in Table 1. A-C) Gene expression (Pparg or its target genes Fabp4 and Plin1) was quantified after 7 days by qRT-PCR. D) Regression analysis of association between Fabp4 expression and lipid accumulation means (data are from Figure 4). The green “X” is Vh. The red “X” is Rosi. r = Pearson’s correlation coefficient. Data are presented as means ± SE from 25 (Vh) or 4–8 (test chemicals) independent bone marrow preparations. Statistically different from Vh-treated (* p<0.05, ** p<0.01, ***p<0.001, ANOVA, Dunnett’s). ATRA-treated was tested separately vs. Vh-treated (* p<0.05, Student’s t test).

To determine whether activation of adipogenesis was associated with suppression of osteogenesis, effects on alkaline phosphatase activity and mRNA expression of osteogenic genes were determined for a subset of the ToxPi chemicals. These chemicals were chosen based on inducing a statistically significant increase or decrease in lipid accumulation in any one of the three models, and they induced a spectrum of effects on lipid accumulation in BM-MSCs (decreasing, not changing, and moderately or strongly increasing). Alkaline phosphatase is a widely recognized biomarker of osteoblast activity and is involved in skeletal mineralization (Sabokbar et al., 1994). As expected, alkaline phosphatase activity was significantly decreased by the positive controls Rosi and TBT (Figure 6A). Of the ToxPi chemicals, only Piogl significantly reduced alkaline phosphatase activity (Figure 6B). However, there was a strong association between increasing lipid accumulation and decreasing alkaline phosphatase activity (Figure 6C), as would be expected if adipogenesis and osteogenesis are co-regulated.

Figure 6. Suppression of alkaline phosphatase activity by test compounds.

Primary bone marrow cultures were established and differentiation was initiated as described in Figure 4. Alkaline phosphatase activity was quantified after 12 days. A) Positive controls. B) Test compounds. Hatched bars = Vh. Light gray bars = negative controls. Dark gray bars = experimental compounds. C) Regression analysis of association between alkaline phosphatase activity and lipid accumulation means. The green “X” is Vh. The red “X” is Rosi. r = Pearson’s correlation coefficient. Data are presented as means ± SE from 4 independent bone marrow preparations. Statistically different from Vh-treated (** p<0.01, *** p<0.001, ANOVA, Dunnett’s).

Changes in gene expression appeared to be a more sensitive indicator of the potential for suspected adipogens to suppress osteogenesis. Runx2 is the master regulator of osteogenesis and controls the expression of osterix (Osx), which also is an essential osteogenic transcription factor. Osteocalcin (bone gamma-carboxyglutamate protein, Bglap) is expressed by late osteoblasts and is involved in mineralization (Bonewald, 2011). Runx2 expression was significantly suppressed only by Rosi, FenHy, Piogl and TBBPA (Figure 7A, Figure S2B), Osx and Bglap expression were suppressed by Allet, FenHy, Piogl, Prall, Quino and TBBPA (Figures 7B-C). That the suppression of gene expression occurred as a result of a specific biological process and not overt toxicity is supported by the fact that Rn18s expression was not suppressed by any treatment, relative to Vh controls (data not shown). ATRA also suppressed Osx and Bglap expression. There was a non-linear relationship between PPARγ-mediated gene expression (Fabp4) and Runx2-mediated gene expression (Osx)(Figure 7D).

Figure 7. Suppression of osteogenic gene expression by test compounds.

Primary bone marrow cultures were established and differentiation was initiated as described in Figure 5. A-C) Gene expression (Runx2 or its target genes Osx and Bglap) was quantified after 7 days by qRT-PCR. D) Regression analysis of association between Osx and Fapb4 (data are from Figure 5) expression means. The green “X” is Vh. The red “X” is Rosi. r = Pearson’s correlation coefficient. Data are presented as means ± SE from 24 (Vh) or 4–8 (test chemicals) independent bone marrow preparations. Statistically different from Vh-treated (* p<0.05, ** p<0.01, ***p<0.001, ANOVA, Dunnett’s).

4. Discussion

We need to know not only what chemicals in the environment and in consumer products are acting as adipogens but also how their interaction with PPARγ, a molecular mediator common to adipose and bone homeostasis (Tontonoz et al., 1994; Lecka-Czernik et al., 1999), results in a pathological phenotype. Recent reviews have outlined the different models that are used to identify and study adipogens (Ruiz-Ojeda et al., 2016; Chamorro-Garcia and Blumberg, 2019). Here, we directly compare 3T3-L1 cells, OP9 cells and BM-MSCs, and the results suggest that a multi-model approach is needed to identify both direct activators of adipogenesis (i.e., PPARγ ligands) and indirect activators (i.e., RXR ligands), as well to investigate coordinate effects on adipogenesis and osteogenesis.

3T3-L1 cells are the most used model in the study of adipogens (Green and Kehinde, 1975; Ruiz-Ojeda et al., 2016). A significant drawback to this model, however, is the variability of the efficacy of differentiation, which has been shown to depend on the commercial source, passage number and culture dishes used (Kassotis et al., 2017). Known PPARγ ligands (rosiglitazone, pioglitazone, tetrabromobisphenol A and triphenyl phosphate) acted as expected, inducing adipocyte differentiation and lipid accumulation, and the known retinoic acid receptor agonist, retinoic acid, decreased adipocyte differentiation. Two limitations of our 3T3-L1 experiments should be noted. First, the concentration of dexamethasone used was 25 nM, a concentration above the EC₅₀ for activation of the glucocorticoid receptor in 3T3-L1 cells by dexamethasone (3 nM)(Nadaka et al., 1987). This may result in false negatives, as subsequent addition of a test chemical that also is a glucocorticoid agonist would could not increase an already maximal receptor activation. Second, we unexpectedly observed that the glucocorticoid receptor ligands, dexamethasone and corticosterone, decreased lipid accumulation. A potential explanation is that the glucocorticoids were included throughout the 10 day protocol. If fully mature adipocytes were present, treatment of mature adipocytes with glucocorticoids increases lipolysis, resulting in a decrease in lipid content of adipocytes (Campbell et al., 2011). The data from this study show that 3T3-L1 cells respond to those potential adipogens that activate PPARγ transcriptional activity. Given the critical role of PPARγ in regulating adipocyte differentiation and function (Tontonoz et al., 1994; Imai et al., 2004), it is not surprising that strong “hits” on the PPARγ ToxPi targets were predictive of adipogenic activity. Quinoxyfen also was shown to support adipogenesis in two other analyses of ToxCast/ToxPi predictions of adipogenic potential (Janesick et al., 2016; Foley et al., 2017). Chemicals of the allethrin class were confirmed as adipogenic in our studies and studies using human adipose-derived stem cells (Foley et al., 2017). Fenthion caused increased body weight in two studies summarized in the EPA ToxRef database (Kowalski et al., 1989; Leser and Suberg, 1990).

An important feature of 3T3-L1 cells is their ability to differentiate into both white and brite adipocytes (Vernochet et al., 2009). The white adipogenic, brite/brown adipogenic and insulin sensitizing activities of PPARγ are regulated separately through differential post-translational modifications and co-regulator recruitment, with ligands having distinct abilities to activate each function of PPARγ (Choi et al., 2010; Choi et al., 2011; Qiang et al., 2012; Villanueva et al., 2013). Along these lines, we recently showed that the environmental adipogens tributyltin and triphenyl phosphate, while adipogenic, have limited capacity to induced brite adipogenesis and mitochondrial biogenesis (Kim et al., 2018; Kim et al., 2020). Thus, while 3T3-L1 cells may not be the most sensitive and consistent model for identifying novel adipogens, they are important for mechanistic studies and investigating the potential for adipogens to induce white versus brite adipogenesis.

OP9 cells are less commonly used to study adipogenesis. The most significant drawback of this model is that they do not appear to be able to differentiate into brite adipocytes (data not shown). This cell line has two significant advantages. First, OP9 are significantly less finicky than 3T3-L1 cells in terms of the experimental conditions required for efficient adipocyte differentiation, likely because they are further along in commitment to adipogenesis than 3T3-L1 cells (Wolins et al., 2006; Kassotis et al., 2017). While in this study, we used the same differentiation protocol for both 3T3-L1 and OP9 cells, we have found that insulin (167 nM) alone is sufficient to support ligand-induced differentiation (data not shown). Second, RXR ligands were significantly more efficacious at stimulating adipogenesis in OP9 cells than in either 3T3 L1 or primary BM-MSCs. We have observed this previously and concluded that it was not a result of differential nuclear receptor expression (Kassotis et al., 2017). In order for an RXR ligand to induce adipogenesis, RXR must be able to permissively activate PPARγ (Kojetin et al., 2015), and this permissivity depends upon the shape that the receptors assume upon RXR ligand binding as well as its ability to recruit coregulators (Schulman et al., 1997; Schulman et al., 1998; Shao et al., 2000). Thus, we hypothesize that the difference in ability of RXR ligands to induce adipogenesis results from the coregulator milieu present in a given adipocyte precursor cell type. As such, OP9 cells are highly sensitive at responding to both direct and indirect activation of PPARγ and are important for studies investigating the effects of RXR ligands on adipogenesis.

Primary bone marrow MSCs are the least commonly used model for studying adipogens. The most important drawback is that they are more challenging and expensive to obtain (either deriving them from mice or obtaining human MSCs commercially). The significant advantage of primary MSC models is that they are multi-potent and can be used to study both adipogenesis and osteogenesis. Human exposure to the therapeutic PPARγ ligand rosiglitazone is known to increase risk of fracture, particularly in post-menopausal women (Aubert et al., 2010; Bilik et al., 2010; Schwartz et al., 2006). In the process of stimulating adipocyte differentiation, rosiglitazone diverts differentiation away from osteogenesis (Lecka-Czernik et al., 1999). We have shown that tributyltin, triphenyltin, triphenyl phosphate, and mono-(2-ethylhexyl) phthalate, all environmental PPARγ activators, also divert MSC differentiation away from osteogenesis (Pillai et al., 2014; Watt and Schlezinger, 2015). Thus, it is not surprising that the data here revealed a strong relationship between potential to induce adipocyte differentiation and to suppress osteoblast differentiation. The efficacy of adipocyte differentiation was low relative to the 3T3-L1 and OP9 models, but it is likely to be higher when the differentiation protocol is solely designed to stimulate adipocyte differentiation (e.g., including IBMX, dexamethasone, and insulin) rather than allowing for both adipocyte and osteocyte differentiation (e.g., including ascorbate, β-glycerol phosphate, dexamethasone, and insulin), as we did here. While we used mouse BM-MSCs in this study, human adipose-derived MSCs have been shown to be responsive to environmental PPARγ ligands, may be more responsive to RXR ligands than mouse-derived MSCs, and will generate white and brite adipocytes (Kirchner et al., 2010; Kim et al., 2020); however, the differentiation protocol is the most complex (Lee and Fried, 2014).

An important caveat of this work is that any effects of these chemicals on humans will depend on levels and timing of exposure, as well as on pharmacokinetics. The in vitro assays used here have no or limited xenobiotic metabolism capacity, so we have only been able to test the effects on the parent compounds. Research is underway to be integrate metabolic activity with generic in vitro assays (DeGroot et al., 2018), and such testing is an obvious follow-up to the current study. Additionally, one can put the adipogenic effects seen here into a dose context by calculating the oral equivalent dose that would lead to the observed in vitro effects using in vitro toxicokinetic models (Wambaugh et al., 2018), and the comparing with predicted or known exposure levels.

Use of in vitro models is an important first step in identifying chemicals that can enhance adipogenesis. No single model can address the multiple facets of PPARγ activation (direct versus indirect activation), adipocyte phenotype (energy storing white adipocyte differentiation versus energy dissipating brite adipocyte differentiation), or downstream outcome (coordinate suppression of osteocyte differentiation. The results of this study show that investigations in multiple in vitro models can provide a more robust analysis of new adipogens and aid in prioritizing testing in vivo.

Highlights.

• Experimental adipogens enhance lipid accumulation in 3T3-L1, OP9 and bone marrow multipotent stromal cells.

• OP9 cells are more sensitive to adipogens that activate RXR than 3T3-L1 cells.

• Adipogens concomitantly decrease osteogenesis as they enhance adipogenesis in bone marrow multipotent stromal cells.

1. Abbreviations:

Bglap

bone gamma-carboxyglutamate protein

BM-MSC

bone marrow MSC

DMSO

dimethyl sulfoxide

Fabp4

fatty acid binding protein 4

FBS

fetal bovine serum

LG268

LG10026

MDC

metabolism disrupting chemical

MSC

multipotent stromal cell

Osx

osterix

PBS

phosphate buffered saline

Plin1

perilipin 1

pNPP

p-nitrophenyl phosphate

PPAR

peroxisome proliferator activated receptor

RFU

relative fluorescence unit

Rosi

rosiglitazone

Runx2

Runt-related transcription factor 2

RXR

retinoid X receptor

SE

standard error

TBT

tributyltin

TPhP

triphenyl phosphate

ToxPi

Toxicological Priority Index

Vh

vehicle