Culture of HEPM cell line

Human embryonic palatal mesenchyme (HEPM) cells, available commercially (ATCC, CRL-1486), were cultured in DMEM media, with high concentration of glucose and pyruvate, supplemented with 10% fetal bovine serum and penicillin/streptomycin. The HEPM cells were kept at low passage numbers and inspected visually for signs of differentiation. Topiramate (Sigma-Aldrich, St. Louis, MO) was resuspended in ethanol. HEPM cells were treated with 1000μM topiramate and cultured at 37°C for 6 hours. Cells were briefly washed with PBS, scraped and flash-frozen for subsequent RNA and protein extraction.

Isolation and culture of primary MEPM cells

Palatal shelves were excised from wildtype E13.5 mouse embryos. Briefly, the embryo heads were cut, followed by removal of the lower jaw and tongue, thus exposing the palatal shelves. The palatal shelves were subsequently dissected, and the cells dissociated into a single-cell suspension by incubating the palatal shelves in 0.25% Trypsin for 10 minutes and then vigorously pipetting up and down to mechanically separate the cells. The resulting Mouse Embryonic Palatal Mesenchyme (MEPM) cells were then cultured in Dulbecco’s Modified Eagle Medium (DMEM) with high glucose and supplemented with 10% Fetal Bovine Serum and Penicillin/Streptomycin. MEPMs were used fresh, never-passaged, before drug treatment. MEPM cells were treated with 25, 50 or 100 μM topiramate (APExBio, Houston, TX) and incubated for 6 hours at 37°C. GABA inhibitor (10 μM), flumazenil (APExBio, Houston, TX), was added with 50 μM topiramate for the 6 hour treatment duration. Ethanol was used for vehicle treatment. All experiments involving animals were carried out with a protocol approved by the University of Kansas Medical Center (KUMC) Institutional Animal Care and Use Committee, in accordance with their guidelines and regulations (Protocol Number: 2018–2447). The experiments reported in this study are from 12 timed-pregnant C57BL/6J female mice (The Jackson Laboratory, Bar Harbor, ME) that were euthanized using a CO₂ chamber followed by harvesting of E13.5 embryos.

Antibody array analysis

Protein extracts from HEPM cells with vehicle (Control) or with topiramate treatment (1000 μM for 6 hours) were analyzed by Full Moon BioSystems (Sunnyvale, CA) Cell Signaling Explorer antibody arrays (SET100), according to manufacturer’s protocol. Briefly, control and topiramate-treated HEPM cell lysates were labeled with Cy3 fluorophore using the antibody array assay kit from Full Moon BioSystems (KAS02). Antibody array slides were independently incubated with labeled lysates, washed, and scanned using microarray scanner from Agilent (Santa Clara, CA).

Statistical analysis of antibody array data

Each antibody array comprised of two technical replicates. The experiment was performed in biological duplicates for downstream statistical analysis. Each probe from each of the biological and technical replicates was first determined to be significantly expressed relative to its background intensity. This significance calculation was based on a ≤ 0.05 cutoff of the Benjamini and Hochberg [20] adjusted p-value of the t-statistic of the difference between the mean expression (F532 mean) and mean background (B532 mean) intensities. A gene was regarded as not significantly expressed under a particular treatment if any one of its biological replicates was not significantly expressed. The first filtering step removed all genes that were not significantly expressed in either one of the treatment groups. The remaining probes were background corrected using a modified Robust Multi-array Average (RMA) [21] algorithm for protein arrays. The background adjusted probes were log transformed (base 2) and quantile normalized. Technical replicates were then averaged (geometric average) to give the background adjusted normalized expression for each biological replicate. A two-way ANOVA model (factor 1: treatment with levels Control and topiramate treated; factor 2: antibody array with levels array 1 and array 2) was fitted to the data to determine the gene level significance of the difference in expression between Control and topiramate treated samples. Protein expression with a p-value ≤ 0.05 and absolute fold change ≥ 1.15 were deemed sufficiently differentially expressed for protein arrays, yielding 40 proteins for further analysis.

Gene interaction analysis

The Ingenuity Pathway Analysis software (IPA, Qiagen, www.qiagenbioinformatics.com) was used to build a gene interaction network of the 40 differentially expressed proteins and 107 known genes associated with orofacial clefts identified by IPA. The aim was to establish a single network of interactions between as many of the 40 differentially expressed proteins from the antibody array and the 107 genes associated with orofacial clefts. IPA performs this task with the aid of its knowledge database consisting of literature-based curated information on genes and gene product interactions.

Quantitative PCR analysis

Quantitative PCR (qPCR) analysis was used to test expression of the TGFB1 in HEPM cells treated with 1mM Topiramate for 6 hours and Tgfb1 in MEPM cells treated with 50μM Topiramate for 6 hours. RNA was extracted using NucleoSpin RNA XS kit (Takara, Kusatsu, Japan). 1 μg of RNA from each sample was used to generate cDNA with qScript cDNA SuperMix (QuantaBio, Beverly, MA). OAZ1 and B2m housekeeping genes were used to normalize data for HEPM and MEPM cells, respectively. Analysis was performed on 4–5 sets of biological replicates, each with 2 technical replicates per gene. Statistical significance was calculated using a Student’s t-test. Primer sequences are listed in S1 Fig.

Western blotting

For protein extraction, MEPM cells were briefly washed with PBS, scraped and either flash-frozen or lysed immediately. Cells were lysed by suspension in radioimmunoprecipitation assay (RIPA) buffer with HALT protease inhibitor Cocktail (Thermo Scientific, Waltham, MA) and by agitation for 30 minutes at 4°C. Cell lysates were centrifuged for 10 minutes at 13,000 rcf and the protein extracts (supernatant) collected. Lysates were then electrophoresed in 4–15% gradient Mini-Protean TGX Stain-Free precast gels (Bio-RAD, Hercules, CA). After electrophoresis, the gels were exposed to UV light for 45 seconds to develop the total protein signal and imaged on a ChemiDoc System (Bio-RAD, Hercules, CA) before being transferred onto Immobilon PVDF membranes (EMD Millipore, Billerica, MA). PVDF membranes were then blocked in Odyssey Blocking Buffer (Li-Cor, Lincoln, NE) either overnight at 4°C or at room temperature for 1 hour. Primary antibodies used were anti-TGFβ1 (1:1000; Abcepta, AP12348A, Cambridge, MA), anti-phospho-SMAD2 (1:5000; Cell Signaling Technologies, 3108, Danvers, MA), anti-SMAD2 (1:5000; Cell Signaling Technologies, 5339, Danvers, MA) and anti-SOX9 (1:5000; Abcepta, AM1964b, San Diego, CA), and anti-SOX10 (1:5000; Aviva Systems, ARP33326, San Diego, CA). Secondary antibodies used were HRP-linked goat anti-rabbit IgG (1:10,000; Cell Signaling Technologies, Danvers, MA) and HRP-linked goat anti-mouse IgG (1:10,000; Santa Cruz Biotechnologies, Dallas, TX). Femto SuperSignal West ECL reagent (Thermo Scientific, Waltham, MA) was used to develop the signal. Image Lab software (Bio Rad, Hercules, CA) was used to quantitate total protein and western blot intensity. Each blot was normalized to the total protein loaded, and then fold change calculated by dividing total drug-treated samples by vehicle-treated sample.

Immunofluorescence and imaging analysis

MEPM cells, cultured as described above, were fixed in 4% paraformaldehyde (PFA) for 10 min, blocked in phosphate buffered saline (PBS) with 1% goat serum and 0.1% Tween, and stained using Anti-TGFβ1 (1:1000; Abcam, Cambridge, MA). After staining, coverslips were mounted in containing DAPI (Vector Labs, Burlingame, CA). Individual cells were imaged, and the levels of TGFB1 fluorescence were quantitated in at least 30 cells per treatment from 3 independent experiments using NIH ImageJ software. Briefly, we used NIH ImageJ to calculate the corrected total cell fluorescence (CTCF) in each cell, using the formula: CTCF = Integrated Density–(Area of selected cell x Mean fluorescence of background readings).

Antibody-array-based analysis of HEPM cells following topiramate treatment

Protein extracts from HEPM cells with supra-physiological topiramate treatment (1000 μM for 6 hours) or without the treatment (Control) were assayed by Full Moon BioSystems (Sunnyvale, CA) Cell Signaling Explorer antibody-array. The Cell Signaling Explorer array includes antibodies for 1358 individual proteins, in two technical replicates, encompassing 20 cellular pathways. The antibody array experiment was performed with two biological replicates. The results were analyzed for statistical significance as described in the Materials and Methods section. Protein levels of 57 gene products were significantly altered (p<0.05, S1 Table). We used a 1.15-fold cutoff (|FC|≥1.15) [22, 23], resulting in 40 differentially expressed proteins with 19 proteins downregulated and 21 upregulated (Table 1). To determine the importance of these 40 altered gene products to orofacial morphogenesis and OFCs, we used Ingenuity Systems pathway analysis (IPA) as well as manual literature curation.

Table 1

List of 40 proteins with statistically significant change in expression and a 1.15-fold-change cut-off in HEPM cells following topiramate treatment.

Protein Name	Symbol	UniProtKB ID	Antibody Array ID	Fold Change
Checkpoint kinase 2	CHEK2	O96017	659	1.96
Low-density lipoprotein receptor class A domain-containing protein 1	LDLRAD1 / LRP1	Q5T700	109	1.64
Aldehyde dehydrogenase 7	ALDH3B1	P43353	92	1.46
NADPH oxidase 5	NOX5	Q96PH1	282	1.46
Proto-oncogene tyrosine-protein kinase receptor Ret	RET	P07949	490	1.42
Cardiac troponin I	TNNI3(cTnl)	P19429	146	1.39
Transforming Growth Factor beta 1	TGFB1	P01137	377	1.39
Unconventional myosin Id	MYO1D	O94832	276	1.37
Tyrosine-protein kinase receptor 3	TYRO3	Q06418	478	1.33
Cyclosome 1	APC1	Q9H1A4	1093	1.32
Superoxide dismutase 1	SOD1	P00441	996	1.31
Oral cancer-overexpressed protein 1	ORAV1 / ORAOV1	Q8WV07	1078	1.30
Growth arrest and DNA damage-inducible proteins-interacting protein 1	GADD45GIP1	Q8TAE8	1285	1.27
Phosphatidylinositol-glycan biosynthesis class H protein	PIGH	Q14442	453	1.25
CD30 Ligand	CD153	P32971	220	1.24
Patched	PTCH1	Q13635	35	1.23
Cadherin-10, T2-cadherin	CDH10	Q9Y6N8	415	1.21
ATP synthase subunit delta, mitochondrial	ATP5D	P30049	261	1.19
TNF Receptor1-associated DEATH domain protein	TRADD /TNFR1	Q15628	716	1.19
Alcohol dehydrogenase class 4 mu/sigma chain	ADH7	P40394	91	1.19
Interferon Regulatory Factor 4	IRF4	Q15306	730	1.18
Bcl10-interacting CARD protein	C9ORF89	Q96LW7	422	-1.15
S-phase kinase-associated protein 1	SKP1A/p19	P63208	363	-1.16
B melanoma antigen 3/ Cancer/testis antigen 2.3	BAGE3	Q86Y29	420	-1.20
Cancer-associated Gene Protein	ATBP3 / MTUS1	Q7Z7A3	907	-1.20
SUMO1-activating enzyme 2	UBA2	Q9UBT2	874	-1.22
Tissue inhibitor of metalloproteinases 2	TIMP2	P16035	537	-1.22
RING finger and WD repeat domain protein 2	RFWD2	Q8NHY2	1107	-1.23
Coagulation factor VII (light chain, Cleaved-Arg212)	FA7	P08709	1156	-1.29
Rho guanine nucleotide exchange factor 3	ARHGEF3	Q9NR81	1123	-1.30
Cytochrome c oxidase assembly protein COX11	COX11	Q9Y6N1	911	-1.31
Diacylglycerol Kinase eta	DGKH	Q86XP1	908	-1.36
ATP-binding cassette sub-family A member 8	ABCA8	O94911	848	-1.37
Shugoshin-like 1	SGOL1	Q5FBB7	895	-1.37
Guanylate Cyclase Beta 1	GUCY1B3	Q02153	853	-1.38
Cell division cycle 7 related kinase	CDC7	O00311	1044	-1.38
Aldo-keto reductase family 1 member C-like protein 1	AKR1CL1	Q5T2L2	246	-1.41
Interleukin 2	IL2	P60568	832	-1.55
Cytochrome P450 7B1	CYP7B1	O75881	802	-1.82
Ubiquitin-protein ligase E3B	UBE3B	Q7Z3V4	291	-2.19

IPA analysis of proteins with altered expression following topiramate treatment shows connectivity with genes associated with orofacial clefts

The antibody array analysis identified statistically significant changes in expression of 40 proteins following topiramate treatment (Table 1; p<0.05, |FC|≥1.15). Separately, IPA software reported 107 known genes associated with orofacial clefts (S2 Table). Only 18 of the 107 OFC genes were represented on the Full Moon BioSystems Cell Signaling Explorer antibody array. Among these 18 proteins encoded by OFC genes, only TGFβ1 was significantly altered (increased) upon topiramate treatment of HEPM cells.

We also analyzed the 40 gene products with IPA for changes in diseases and bio-functions (S3 Table). The top predicted categories were related to “cell death and survival”, and “organismal growth and development”. Overall, IPA predicted that topiramate treatment decreased cell viability (S3 Table; IPA Ranks 13, 29) and increased apoptosis (S3 Table; IPA Rank 37). However, IPA also correctly predicted increased cell viability of neurons (S3 Table; IPA Rank 7) [24], consistent with a cell-type specific effect of topiramate. The top IPA predicted networks also indicated an effect on cell growth and survival pathways (S4 Table). Next, we looked at connectivity of the genes encoding the 40 altered proteins to the IPA-reported 107 genes associated with orofacial clefts (Fig 1). TGFβ1 was the only OFC gene among the 40 topiramate-altered proteins in our analysis. A single network was able to connect 22 (9 downregulated, 13 upregulated) of these 40 (55%) genes encoding differentially expressed proteins to 87 OFC genes (81%), either directly or indirectly. TGFβ1 showed the highest number of connections to known genes associated with orofacial clefts (Fig 1). Therefore, we focused our validation studies on TGFβ1.

Fig 1

TGFβ1 showed highest connectivity in IPA-generated network of differentially expressed gene products from topiramate-treated HEPM cells and known orofacial clefting-associated genes.

The 40 gene products with differential expression following 6-hour, 1000μM topiramate treatment of HEPM cells were analyzed with Ingenuity Systems pathway analysis (IPA) software for possible interaction with 107 IPA-identified genes associated with orofacial clefts. A single resulting network accounted for 22 (55%) of the topiramate-treated HEPM genes in association with 87 (81%) known OFC-related genes. The upregulated or downregulated genes from the HEPM data were colored blue or green, respectively. The only gene common between the two datasets was TGFB1, which also displays the highest connectivity in the network.

Validation of TGFβ1 upregulation in primary Mouse Embryonic Palate Mesenchyme (MEPM) cells following treatment with physiological levels of topiramate

Our analysis indicated TGFβ1 as the central molecule affected by topiramate treatment of HEPM cells. To confirm that our finding was not an artifact of using an immortalized cell line and supraphysiological levels of topiramate, we used primary MEPM cells and physiological topiramate doses. The peak serum level for even topiramate monotherapy is approximately 25–50 μM [25–27]. Therefore, we initially included both the physiological 50 μM concentration and two supraphysiological concentrations to treat fresh primary MEPM cells from E13.5 embryonic palates for 6 hours (Fig 2A–2E). Following treatment, cells were fixed and immunostained using an anti-TGFB1 antibody (Fig 2A–2D). Individual cells were imaged, and the levels of TGFB1 were quantitated using NIH ImageJ software. The experiment was repeated three times with different primary MEPM isolations. We saw a significant increase in intracellular TGFβ1 level at the physiological 50μM topiramate treatment (Fig 2E).

Fig 2

Topiramate treatment of primary MEPM cells upregulated TGFβ1 expression via GABA receptors.

To validate upregulation of TGFβ1, we isolated primary mouse embryonic palate mesenchyme (MEPM) cells from E13.5 embryos and treated them with 25μM, 50μM, 250μM or 1mM topiramate for 6 hours, as indicated. These cells were analyzed for TGFβ1 expression by immunostaining (A-E) and Western blotting (F-G). There was a significant increase in intracellular TGFβ1 expression upon as little as 50μM topiramate treatment (A-E), which was quantitated in at least 30 cells per treatment from 3 independent experiments (**, p<0.003; ***, p<0.0003). The scale bar represents 50μm (A-D). Western blot analysis also showed a significant increase upon treatment with 50μM topiramate (F, G). We also showed increased phospho-SMAD2 (P-SMAD2) levels with both 25μM and 50 μM topiramate treatments (F, G), indicating an upregulation of TGFβ1 signaling cascade (*, p<0.019). No significant changes were observed in total SMAD-2 level upon 50 μM topiramate treatment (F, G). The upregulation of TGFβ1 protein expression upon 50 μM topiramate treatment is blunted in the presence of 10 μM GABA receptor inhibitor, Flumazenil (F, G).

We also performed Western blot analysis following 25μM and 50μM topiramate treatments (Fig 2F) and showed a significant increase with 50μM topiramate (Fig 2G). There was no increase in TGFB1 (HEPM) or Tgfb1 (MEPM) mRNA transcripts at the end of the 6-hour topiramate treatment (S1 Fig). To test for an upregulation of the TGFβ1 signaling pathway, we also assessed levels of phospho-SMAD2 (P-SMAD2), a downstream effector molecule (transcription factor) of the ligand-bound TGFβ receptor that requires phosphorylation to translocate to the nucleus. Indeed, P-SMAD2 levels were increased following both 25μM and 50μM topiramate treatment (Fig 2F and 2G), indicating a strong upregulation of TGFβ signaling via SMAD-dependent pathway.

Topiramate treatment increased TGFβ1 protein level via GABA receptors

Since topiramate has been reported to upregulate GABA levels and GABA_A receptor-based signaling [28], which in turn has been shown in other systems to increase TGFβ levels [29, 30], we wanted to see if the same was true in MEPM cells. Thus, we treated the MEPM cells with 50μM topiramate in the presence of 10μM flumazenil, a selective inhibitor of GABA_A receptors [31]. Indeed, flumazenil was able to blunt the effect of topiramate on TGFβ1 level (Fig 2F and 2G), indicating that topiramate was mediating its effect through GABA_A receptors.

Topiramate treatment of primary MEPM cells results in increased SOX9 expression

After validating the involvement of TGFβ signaling, we wanted to look at a TGFβ target gene with a role in orofacial clefting. We decided to look at expression of SOX9, a transcription factor belonging to the SOXE group with a key role in regulating chondrocyte function [32, 33]. SOX9 mutations have been identified in patients with Pierre Robin sequence [34] and campomelic (or acampomelic) dysplasia with or without sex reversal [35, 36]. Pierre Robin sequence is a series of defects including small jaw, a posteriorly placed tongue and cleft palate [34]. Campomelic dysplasia is an autosomal dominant skeletal disorder that is characterized by bowed limbs, hypoplastic or hypomineralized bones, and small chest size. Cleft palate, micrognathia (including Pierre Robin sequence), flat face and hypertelorism are also frequent features of Campomelic dysplasia. Interestingly, SOX9 mutations in patients with Pierre Robin sequence were located in the regulatory region affecting SOX9 expression. Furthermore, both knockdown and overexpression of Sox9 in mouse has been shown to result in cleft palate phenotypes [32, 33, 37]. Therefore, an effect on SOX9 expression downstream of TGFβ1 in our system would represent a plausible pathogenetic mechanism of topiramate-based facial clefts. Indeed, we found that SOX9 expression is increased following topiramate treatment (Fig 3A and 3B), which is consistent with a role for TGFβ1 in stabilizing SOX9 protein [38]. Expression of another SOXE transcription factor group member associated with orofacial clefts, SOX10, is not altered upon topiramate treatment (Fig 3A and 3C), showing specificity for SOX9 upregulation. Our results with topiramate suggest that TGFβ1-mediated altered expression of SOX9 underlies the teratogenicity of this antiepileptic drug.

Fig 3

Topiramate treatment of primary MEPM cells resulted in increased SOX9 expression.

We looked at expression of SOX9, a TGFβ1 target gene involved in orofacial clefting. Western blot analysis of topiramate-treated primary MEPM cells resulted in statistically significant increase in SOX9 expression (A,B; **, p<0.006). In contrast, expression of SOX10 is not altered upon topiramate treatment (A,C; ns, not significant).