Effects of multi-walled carbon nanotubes on message and Micro-RNA in human lung BEAS-2B cells

Abstract

Multi-walled Carbon nanotubes (MWCNTs) lack sufficient quality cytotoxicity, toxicity, genotoxicity and genomic data on which to make environmental and regulatory decisions. Therefore, we did a multidisciplinary in vitro study of 3 MWCNTs in human lung cells (BEAS-2B) with the following endpoints: cytotoxicity, DNA damage, reactive oxygen and nitrogen species, lipid peroxidation and mRNA and microRNA expression analyses. The MWCNTs were either unfunctionalized or functionalized with either -OH or -COOH. Doses studied ranged from 0.3 to 100 ug/ml and were exposed to a human lung cell line in vitro for 72 h., with genomic studies being done from 30 ug/ml downward. Some of the genomic pathways that were altered by MWCNT exposure were NRF2 mediated oxidative stress response, DNA damage repair, nuclear excision repair, base excision repair, mitochondrial dysfunction, oxidative phosphorylation, HIF1α signaling, unfolded protein response, protein ubiquitination, ferroptosis and sirtuin signaling pathways. The data suggested that OH functionalized MWCNT caused more and larger gene/microRNA changes, followed by COOH functionalized MWCNT and unfunctionalized MWCNT being the least biologically active. From microRNA target filter analysis, there were altered signaling hubs. MYC is the only hub that altered by all 3 MWCNTs. Signaling hubs that are common to OH and COOH functionalized MWCNTs are GRB2, AR, TP63 and AGO2. The signaling hubs that were only present in OH functionalized MWCNTs are TP53, STAT3 and BRCA1. These signaling pathways and hubs we found in vitro correlated well with the published in vivo pathological effects like oxidative stress DNA damage, inflammation and cancer in MWCNTs treated mice.

Graphical Abstract

Treatment of BEAS-2B cells with 55-OH-MWCNT (multi-walled carbon nanotubes functionalized with -OH) caused many expression changes in mRNA and miRNA. The graph below shows the connected networks between different signaling hubs.

1. INTRODUCTION

Carbon nanotubes (CNTs) are large carbon-based tubular structures that are long and thin, between 1 −100 nm in diameter, and can be hundreds to thousands of nanometers long. Nanotubes are a hundred times stronger than steel and one-sixth its weight. The properties of the CNTs depend on the diameter and length of the tubes, and their overall nanostructures and the presence or absence of non-carbon atoms and functional groups. Single-walled carbon nanotubes consist of one layer of a rolled-up graphene sheet, while multi-walled carbon nanotubes (MWCNTs) consist of multiple rolled layers of graphene sheets. CNTs have high mechanical strength, flexibility and impressive thermal and electrical conductivity and thus make them attractive for various applications [1]. However, because of the chemical structure, it is very difficult to dissolve or disperse CNTs into water or other liquid solutions, and is very difficult to get CNTs to physically interact with other chemicals [2]. In order to circumvent these two drawbacks, functionalization has been employed to improve CNT’s surface properties and reactivity.

The production of CNTs exceed several thousand tons per year. There are wide applications for CNTs, they are used in energy storage, automotive parts, semi-conductors, boat hulls, sporting goods, thin-film electronics and medical applications. CNTs are being highly researched as efficient drug delivery and biosensing methods for disease treatment and health monitoring. Humans are exposed CNTs through inhalation, dermal exposure, and oral exposure through contaminated food and water.

Despite the huge amount of interest and investment in MWCNTs, the toxicological and genomic impacts of these and other engineered nanoparticles are not well understood. MWCNTs are similar to asbestos in the physical dimensions and bio-persistence and MWCNTs were found to exhibit asbestos-like pathogenicity [3]. Data from animal studies showed that MWCNTs induced sustained inflammation, DNA damage, fibrosis and lung cancer following long-term inhalation [4–6], and the formation of mesothelioma when injected into the peritoneal cavity of susceptible mice [7]. Recent studies showed that when MWCNTs were administered to TP53 heterozygous mice or rats intraperitoneally, both animal models developed mesothelioma in a dose dependent manner [8, 9]. However, the mechanisms seemed to be different from asbestos [10]. Published data from in vitro toxicity studies showed that MWCNTs with smaller diameter demonstrated greater toxicity than those with larger diameter [11], and functionalized MWCNTs are more genotoxic, but less cytotoxic than their unfunctionalized counterparts [12]. However, the in vitro adverse effects to MWCNTs, especially functionalized MWCNT, are mostly unknown.

The most common routes of nanomaterial exposures are by inhalation, ingestion and absorption through skin [13]. Therefore, in this genomic and miRNA study, we examined the effects of a MWCNT and its functionalized (-OH and -COOH) counterparts in human lung epithelial cells (BEAS-2B). The results showed that -OH or -COOH functionalized MWCNTs were more biologically active than the unfunctionalized parent material. The functionalized MWCNTs induced more changes in gene/microRNAs and more signaling pathways (from mRNA IPA analysis) and signaling hubs (from Target filter analysis) than the non-functionalized MWCNT.

2. MATERIALS AND METHODS

2.1. Multiwalled Carbon nanotubes (MWCNTs) and their dispersions

MWCNTs were purchased from Cheap Tubes Inc. (Crafton, VT). The physical and chemical properties obtained from the company website were listed in Table 1. MWCNT with outer diameter between 20–30 nm, length of 10 to 30 micrometer (μm) (33-MWCNT), and its functionalized counter parts, -OH and -COOH, (55-OH-MWCNT and 66-COOH-MWCNT, respectively) were weighed into amber glass vials, dispersion solution was added to make 300 μg/ml of MWCNTs. The dispersion solution is the cell culture media LHC-9 (Thermofisher, Waltham MA. Cat #12680013) with 1% F127 (Sigma-Aldrich, St. Louis, MO. Cat# P2443) added to improve dispersion. The MWCNTs in the dispersion medium were then sonicated on ice with a Qsonic^® ultrasonicator (model Q125, Qsonic, Newtown, CT) for 90 min. After sonication, the mixtures were spun at 10,000 RCF for 10 min. The optical density (OD) at 500 nm was taken before and after the centrifugation to determine the percentage dispersion (% dispersion = OD after spun/OD before spun × 100). The actual concentration of the MWCNTs were calculated by multiplying the original concentration (300 μg/ml) with the % dispersion. The treatment volumes were determined from the actual concentration of the dispersed MWCNTs

Table 1.

Chemical and Physical properties of MWCNTs

MWCNT	cat #	Outer diameter (nm)	Inner Diameter (nm)	length (μm)	specific surface area (m2/g)	Functionalization	% of functionalized group	purity wt%
33-MWCNT	sku 030104	20–30	5–10	10–30	110	NONE	na	>95
55-OH-MWCNT	sku 030204	20–30	5–10	10–30	110	-OH	1.6	>95
66-COOH-MWCNT	sku 030304	20–30	5–10	10–30	110	-COOH	1.2	>95

Note: All characterizations were done by Cheap Tubes Inc.

na: not applicable

2.2. Cell culture and treatments

BEAS-2B, an Ad12-SV40-transformed nontumorigenic human bronchial epithelial cell line, were purchased (American Type Culture Collection, Manassas, VA) between passage numbers 46–57. Cells were cultured in LHC-9 medium, a defined medium without fetal bovine serum. Cells were maintained in 150 cm² flasks at 37 °C and 5% CO_2, with about 95% relative humidity. Cells were passaged when ~85% confluent by washing first with HEPES buffered saline solution (HBSS; Lonza, Basel, Switzerland, Cat#CC-5022), then with Versene (Thermofisher, Waltham MA Cat #15040066) and dislodging with TrypLE^™ (Thermofisher #12605010). TrypLE activity was terminated with trypsin neutralizing solution (TNS, Lonza #CC3101). Both TrypLE^™ and TNS were removed by centrifugation. For each endpoint determination, cells were grown in LHC-9 medium at 37°C, 5% CO₂ until they were ~70% confluent after subculturing. Cells were treated in 25 cm² flasks with 0, 0.3, 1, 3, 10 and 30 μg/ml MWCNTs for 72 h for gene expression analysis and cytotoxicity determinations. All treatments were done in at least 5 replicates. Cells were washed twice with PBS before harvesting.

2.3. Cytotoxicity

MWCNT cytotoxicity was measured at the end of 72-h treatments with MTS [3-(4,5-Dimethylthiazol-2-yl)-5-(3-carboymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium], CellTiter 96^® AQueous One solution cell proliferation assay, Promega, Madison, WI, cat# G3582), Alamar Blue (AlamarBlue^™, ThermoFisher, cat#DAL1100) and SRB (Sulforhodamines B assay kit, Abcam, Cambrige, MA, cat# 235935). These methods were performed according to the manufacturers’ recommendations.

2.4. DNA damage, ROS/RNS measurement s and lipid peroxidation

DNA was isolated from BEAS-2B cells by the salt/spermine method [14] where the proteins in the cells were precipitated by adding salt (NaCl) and the DNA was precipitated with spermine tetrahydrochloride . Protein was isolated from the BEAS-2B cells by lysing the cells according to the salt/spermine method and stopping the method prior to the first proteinase K digestion. To determine the generation of Reactive Oxygen Species (ROS) and Reactive Nitrogen Species (RNS) in BEAS-2B cells after MWCNT exposure, a commercial fluorescence kit (Oxiselect; Cell Biolabs, San Diego, CA, cat #STA-347) was used to quantify ROS/RNS in cell lysates. To determine the extent of oxidative damage to DNA after BEAS-2B cells were treated with MWCNT, a commercial ELISA kit for 8-hydroxy-2′-deoxyguanosine (8-oxo-dG) was used (Oxiselect; Cell Biolabs, San Diego, CA Cat# STA-320). Apurinic/apyrimidinic (AP) sites were determined using a commercial colorimetric assay (Cell Biolabs OxiSelect, cat # STA-324). 4-hydroxynonenal (4-HNE)-protein adducts, products of lipid peroxidation, were measured using an ELISA kit from Cell Biolabs (STA-838).

2.5. RNA extractions

Total RNA was extracted from cells using mirVana^™ RNA extraction kits (Thermo Fisher Scientific) following the manufacturer’s protocol. The concentration of the RNA was determined by Nanodrop (Nanodrop 1000, Thermo Scientific, Waltham MA). The integrity of the extracted RNA was assessed in the Bioanalyzer (model 2100, Agilent Santa Clara, CA), and only RNA with the RNA Integrity Number (RIN) greater than 9.8 were used in RNA and small RNA sequencing.

2.6. RNA sequencing

RNA sequencing was performed by the US EPA’s National Health and Environmental Effects Research Laboratory (NHEERL) Genomic Research Core. RNA sequencing was performed as reported previously [15, 16]. In short, total RNA samples (1 μg per sample) were processed on an Apollo324 liquid handling instrument for mRNA selection with Wafergen kit (PrepX polyA mRNA Isolation Kit, 400047) and continued on library prep with PrepX mRNA (PrepX RNA-Seq 48 Library Kit for Illumina by Wafergen, Fremont CA), then PCR amplified for 15 cycles with 24 index primers. The resulting PCR products were cleaned up on the Apollo324 instrument with PCR Cleanup 48 Protocol. The volume of purified libraries consisting of cDNA synthesized from the mRNA in the cells was 10 μl. One microliter each was taken from every library for Qubit quantitation check with Qubit dsDNA HS kit by Invitrogen and Bioanalyzer quality check with High Sensitivity DNA Chips and Reagents by Agilent Technologies. The molar concentration of each library was estimated by using average molecular size from Bioanalyzer data and concentration from Qubit measurement. Then each library was diluted to 2 nM. The diluted libraries were pooled to make the sample pool for the sequencing run and checked by Qubit to confirm the working concentration. The pooled libraries were denatured and diluted according to Illumina NextSeq protocols. The final concentration for sequencing was 3.5 pM + 2% PhiX (a reference sample provided by Illumina) and it was run for 75 cycles SR (Single Read). The sequencing reactions were run in Illumina NextSeq 500 Sequencer.

MicroRNA sequencing was performed according to manufacturer’s suggestions. In short, the total RNA (1.5 μg per sample) was processed on the Apollo324 robot by Wafergen for small RNA library prep (PrepX Small RNA 48 protocol) with the same reagent kit for RNA-Seq without RNAse III digestion step. After 15 cycles PCR, the products were ethanol precipitated and the pellet was resuspended in 30 μl of water. cDNA fragments from 110bp to 180bp (peak at 145bp) were isolated on Pippin Prep 3% agarose gel with marker F by Sage Science (Beverly, MA). The collected 40 μl cDNA solution were concentrated by Eppendorf Vacufuge Plus to about 15 μl and the sample quality was checked by Bioanalyzer High Sensitivity DNA Chips and Reagents. The cDNA concentration was measured by Qubit dsDNA HS kit. The molar concentration of each library was estimated by using average molecular size from Bioanalyzer data and concentration from Qubit measurement. Then each library was diluted to about 2 nM. The diluted libraries were pooled to make the sample pool for sequencing run and checked by Qubit to confirm the working concentration. The pooled libraries were denatured and diluted according to Illmina NextSeq protocols. The final concentrations for sequencing were around 4 pM + 2% PhiX and it was run for 75 cycles SR (Single Read).

2.7. Data Analysis

Sequencing data analysis was conducted as previously described [15]. All FASTQ files generated by NextSeq 500 were first analyzed by FastQC software for data quality assessment. RNA-Seq profiling was quantified using Partek (E/M) annotation model and STAR-2.5.3a aligner, aligned to human genome hg19. Normalization and contrast were performed using DESeq2. MicroRNA Expression was quantified with mirDeep2, normalization and contrast performed using DESeq2 [17, 18] . Differential expressed genes (DEGs) with adjusted p-value ≤0.05 from the Benjamini-Hochberg method [19] for statistical multi-testing correction were obtained from the DESeq2 for mRNA and microRNA data analyses.

2.8. Ingenuity Pathway Analysis (IPA)

Differential expressed gene (DEG) lists from mRNA sequencing were uploaded to IPA. Core analysis was performed to delineate the canonical pathways altered by each treatment. The DEG lists for microRNA were also uploaded into IPA (IPA 2021 July release) and MicroRNA Target Filter Analysis were performed to find interactions between the messenger RNAs and microRNAs for each treatment. The MicroRNA Target Filter Analysis allows one to access both the experimentally validated and predicted mRNA targets from TargetScan, TarBase, miRecords and the Ingenuity Knowledge Base. For the Target Filter Analysis, the confidence levels for microRNA-RNA interaction, in our study, was set to “Experimentally validated”, which is the highest stringency setting. The “Expression Pairing” filter, meaning only includes miRNA with its regulated mRNA changed in opposite directions, was not applied since sometimes the microRNA change the amount of protein produced without changing the levels of mRNA [20]. Once the target mRNAs were found, both the microRNAs and the mRNAs were added to a pathway canvas in IPA, and the biological interactions between these mRNAs and microRNAs were explored by using IPA’s Pathway Explorer.

3. RESULTS AND DISCUSSION

3.1. Cytotoxicity

We tested the three MWCNTs for cytotoxicity at 0, 0.3, 1, 3, 10, 30 and 100 μg/ml. Similar results were obtained from all three methods and only the results from Alamar Blue assays were shown (Fig 1). From all three methods, MTS, Alamar Blue and SBR, there is minimum toxicity with MWCNTs up to 30 μg/ml, which is the highest concentration used for the genomic analysis. At 100 μg/ml, MTS and SRB showed above 80% of the unexposed control values while, the Alamar Blue assay showed about 70% compared to controls.

Figure 1.

Cytotoxicity results from Alamar Blue assay with 3 MWCNTs in BEAS-2B cells. There were three wells for each concentration. The error bars are standard deviations.

Cells were treated in 96-well plates, 6 wells per concentration.

3.2. ROS/RNS, DNA damage and lipid peroxidation

We tested AP sites, 8-OHdG, 4HNE and ROS/RNS. We did not detect any significant DNA damage (AP sites, 8-OHdG) or lipid peroxidation (4HNE) from our MWCNT treated cells. However, we did see elevated ROS/RNS in all three MWCNTs treated cells (Figure 2). None of the three MWCNTs showed linear dose-response curve of elevation of ROS/RNS. But all three MWCNTs exhibited about 2-fold increase of ROS/RNS at 30 μg/ml.

Figure 2.

ROS/RNS assay. Cells treated with different MWCNTs were harvested and assayed according to Materials and Methods. The sample size was 3, and the error bars are standard errors.

3.3. Differentially expressed gene (DEG) lists

Table 2 shows the DEG lists for both RNA and microRNA. There was not a monotonic dose response relationship between concentration used and the number of DEGs observed for both mRNA and microRNA which is consistent with our previous findings and results from other labs [16, 21, 22]. This is consistent with our previous reports on the effects of other nano metals and metal oxides on gene expression [15, 22–25]. Other researchers also have observed this type of non-linear dose-response curve in mRNA studies with nanomaterials [21, 26]. The functionalized MWCNTs (55-OH-MWCNTand 66-COOH-MWCNT) induced more DEGs for both mRNA and microRNA than the unfunctionalized 33-MWCNT.

Table 2.

Numbers of differentially expressed genes or microRNAs following treatment of BEAS-2B cells with MWCNTs.

MWCNT	33-MWCNT					55-OH-MWCNT					66-COOH-MWCNT
conc. (μg/ml)	30	10	3	1	0.3	30	10	3	1	0.3	30	10	3	1	0.3
RNA	2257	770	1376	920	233	6592	4063	3160	454	1540	4695	4154	5007	4322	1740
microRNA	121	90	76	66	62	358	172	19	53	66	231	247	104	93	29

3.4. Canonical pathways altered from IPA analysis

Biological significance of the identified DEGs were determined using IPA. The top 12 enriched canonical pathways (p-value ≤ 0.05) altered at each dose were listed in Table 3; the lower the number in rank, the lower the p-value and higher probabilities.

Table 3.

Top canonical pathways from IPA analysis

To get a better picture of the alterations in canonical pathways, pathways were grouped together according to their functions and shown in Table 4. The ranking is from IPA canonical pathway analysis, #1 is the one with the lowest p-value. Pathways that were in the top 12 (as in Table 3) were highlighted in yellow.

Table 4.

Ranking of canonical pathways

MWCNT name	33-MWCNT					55-OH-MWCNT					66-COOH-MWCNT
concn. (μg/ml)	30	10	3	1	0.3	30	10	3	1	0.3	30	10	3	1	0.3
Stress response
NRF2-mediated Oxidative Stress Response	7	2	16	na	13	118	37	19	125	74	39	32	16	39	107
Mitochondrial Dysfunction	4	3	10	na	na	4	3	4	7	6	3	5	7	7	4
Oxidative Phosphorylation	9	5	8	52	na	5	2	2	5	5	4	3	9	6	2
Unfolded protein response	3	33	67	30	na	17	41	42	19	10	28	29	22	19	na
Protein Ubiquitination Pathway	1	7	12	6	na	2	9	7	29	7	2	2	2	3	12
Endoplamic Reticulum stress pathway	na	na	na	50	na	69	na	na	104	80	na	na	172	271	na
Glutathione Redox Reactions I	151	53	198	na	na	na	na	na	120	na	na	na	na	na	na
Glutathione Redox Reactions II	137	24	na	na	na	156	na	na	na	16	na	na	na	na	na
P 53 signaling	62	na	na	na	na	46	21	38	53	na	27	23	59	13	35
Sirtuin Signaling Pathway	21	8	11	43	na	10	5	5	6	14	10	4	4	5	13
Senescence Pathway	10	19	123	na	na	39	47	53	na	9	53	47	48	15	53
HIF1α Signaling	6	44	na	na	na	45	63	43	na	na	50	42	18	24	45
NER (Nucleotide Excision Repair, Enhanced Pathway)	81	na	66	na	na	23	14	18	22	36	14	12	8	32	21
BER (Base Excision Repair) Pathway	na	47	na	na	31	48	89	75	8	11	40	27	81	95	37
Hepatic Fibrosis / Hepatic Stellate Cell Activation	44	17	43	na	21	120	133	55	138	na	na	73	68	49	104
Protein synthesis
EIF2 Signaling	33	96	112	na	na	1	1	1	20	44	1	1	1	1	1
Regulation of eIF4 and p70S6K Signaling	38	na	na	na	na	11	4	3	82	89	6	7	5	2	6
mTOR Signaling	na	na	na	na	na	16	10	na	156	na	12	9	3	4	5
Cholesterol Biosynthesis
Superpathway of Cholesterol Biosynthesis	na	4	1	1	1	166	60	32	1	1	na	38	53	22	3
Cholesterol Biosynthesis I	na	9	2	2	2	na	64	25	2	2	na	33	42	33	7
Cholesterol Biosynthesis II (via 24,25-dihydrolanosterol)	na	10	3	3	3	na	65	26	3	3	na	34	43	34	8
Cholesterol Biosynthesis III (via Desmosterol)	na	11	4	4	4	na	66	27	4	4	na	35	44	35	9
Inflammation
IL-6	27	118	59	na	na	na	na	na	na	na	na	153	133	119	na
STAT3 signaling	68	na	na	na	na	80	na	na	na	na	na	183	na	na	na
IL-1 signaling	78	na	92	na	na	na	na	63	na	na	87	56	25	31	64
TGF-β signaling	91	66	80	na	na	na	na	180	na	15	100	135	176	213	63
TNFR1 Signaling	na	60	109	na	na	146	134	na	184	na	na	na	na	252	na
Chemokine signaling	126	na	na	na	na	na	na	171	132	na	na	162	181	205	112
Cancer-related
Hereditary Breast Cancer Signaling	138	na	na	17	na	12	23	16	144	57	20	22	19	28	na
HER-2 Signaling in Breast Cancer	30	na	24	na	na	21	16	9	113	60	15	19	19	8	14
Molecular Mechanisms of Cancer	14	na	19	19	na	18	18	34	155	56	18	13	26	17	25
Non-Small Cell Lung Cancer Signaling	141	na	na	na	27	44	108	111	na	na	61	82	66	57	87
Small Cell Lung Cancer Signaling	na	na	na	na	na	47	61	121	na	na	na	134	60	61	na
Cell cycle related
Cell Cycle: G1/S Checkpoint Regulation	75	na	23	na	na	31	28	81	na	na	26	44	65	94	75
Cell Cycle: G2/M DNA Damage Checkpoint Regulation	108	na	30	45	na	15	15	31	40	na	29	26	51	70	31
Cell Cycle Control of Chromosomal Replication	na	na	na	na	na	37	52	64	46	na	48	71	90	98	43

Note: “na” is not significant

3.5. MicroRNA target filter analysis

Results of microRNA target filter analysis for 30 ug/ml of all three MWCNTs are shown in Figures 3 a–c. In each target filter analysis plot, we identified the important nodes of reaction hubs as genes that have more than 15 connections. Names of the hubs are listed in Table 5a, and the microRNA that are involved in that hub are listed in Table 5b. 55-OH-MWCNT caused most signaling pathway changes, 66-COOH-MWCNT is less, and 33-MWCNT caused the least changes (Table 5). When inspected the major hubs, 33-MWCNT only altered cell proliferation (MYC) and cell cycle progression, Cyclin D1 (CCND1). For 55-OH- and 66-COOH-MWCNT, in addition to cell cycle progression and cell proliferation (MYC), caused changes in signaling pathways involved in inflammation [androgen receptor (AR), growth factor receptor-bound protein 2 (GRB2)] and cancer formation [Tumor protein p63 (TP63) and Argonaut 2 (AGO2)]. Signaling hubs that were altered only in 55-OH-MWCNT treated cells were involved in cell proliferation [Signal Transducer and Activator of Transcription 3 (STAT3)], DNA repair and genome stability [Tumor Protein p53 (TP53) and Breast Cancer 1 (BRCA 1)].

Figure 3.

Target filter analysis results from IPA. 3a. 33-MWCNT at 30 μg/ml. Signaling hubs: MYC and CCND1. 3b. 55-OH-MWCNT at 30 μg/ml. Signaling hubs: MYC, AR, GRB2, TP53, AGO2, TP63, STAT3, and BRCA1. 3c. 66-COOH-MWCNT at 30 μg/ml. Central genes in the signaling hubs are bolded and in larger fonts. Signaling hubs: MYC, CCND1, AR, GRB2, TP63 and AGO2. Red means up-regulation, green means down-regulation.

Table 5a.

Signaling hubs and their functions.

HUB	33-MWCNT	55-OH-MWCNT	66-COOH-MWCNT	Function
MYC	down	down	down	oncogene
CCND1	down	na	down	cell proliferation (cell cycle progression)
AR	na	up	up	suppress inflammation
TP63	na	down	down	negatively regulate TP53
ARGO2	na	up	up	promote tumor progression in KRAS driven mouse NSCL
TP53	na	up	na	tumor suppressor (by triggering cell cycle arrest)
STAT3	na	up	na	promote cell proliferation
GRB2	na	up	up	Inflammatory signaling and Endothelial dysfunction
BRCA1	na	down	na	tumor suppressor (maintaining genome stability)

Note: “na” means not significant; “down” means suppressed, “up” means activated.

Table 5b.

MicroRNAs and their functions related to signaling hubs

	33-MWCNT	55-OH-MWCNT	66-COOH-MWCNT	function of the microRNA
MYC
	miR-200–3p	miR-200b-3p	miR-200–3p	cell proliferation
	miR-24–3p	miR-24–3p		cell proliferation
	miR-141–3p	miR-141–3p	miR-141–3p	cell proliferation
	miR-155–5p	miR-155–5p	miR-155–5p	master regulator of inflammation
		miR-7a-5p		tumor suppressor
		miR-34a-5p	miR-34a-5p	tumor suppressor
		miR-185–3p	miR-185–3p	tumor suppressor
		miR-27a-3p		tumor suppressor
			Let-7a-5p	tumor suppressor target CCND1
CCND1
	miR-146a-5p (in)		miR-146a-5p (in)	tumor suppressor, anti-inflammatory
	miR-26a-5p (in)			oncogenic
	miR-16–5p (in)			tumor suppressor
	miR-503–5p (in)			tumor suppressor by induce apoptosis
	miR-128–3p (in)			tumor suppressor
	miR-424–3p			tumor suppressor
			miR-193a-3p	tumor suppressor target KRAS
			miR-34a-5p	tumor suppressor target KRAS
			let-7a-5p	tumor suppressor target KRAS
TP53
		miR-24–3p		tumor suppressor target KRAS
		miR-103–3p		Tumor suppressor
		miR-378a-3p		suppress inflammation by targeting GRB2
		miR-92a-3p		oncogenic promote proliferation
		miR-132–3p		Tumor suppressor
		miR-125b-2–3p		Tumor suppressor
		miR-17–5p		oncogenic promote proliferation
		miR-27a-3p		oncogenic promote proliferation
		miR-29b-3p		promote inflammation
		miR-155–5p		promote inflammation
TP63
		miR-193a-5p	miR-193a-5p	promote inflammation
		miR-574–3p		oncogenic
AR
		miR-501–3p	miR-501–3p	Tumor suppressor
		miR-146a-5p	miR-146a-5p	tumor suppressor, anti-inflammatory
		miR-486–3p	miR-486–3p	tumor suppressor
		miR-744–3p		tumor suppressor
		miR-222–5p		oncogenic target p27
		miR-93–3p		oncogenic
		miR-29–5p		anti-fibrotic
AGO2
		miR-200b-3p	miR-200b-3p	cell proliferation
		miR-501–3p	miR-501–3p	tumor suppressor
		miR-92a-3p	miR-92a-3p	oncogenic, promote proliferation
			miR-221–3p	oncogenic targeting p27
			miR-7a-5p	oncogenic
			miR-10a-5p	over expressed in lung cancer
		miR-130a-3p	miR-130a-3p	tumor suppressor by targeting KLF3
			miR-16a-5p	tumor suppressor
			miR-100–5p	oncogenic
			miR-17–5p (in)	oncogenic
		miR-29b-3p		oncogenic
		miR-96–5p		oncogenic
		miR-196a-5p		oncogenic
		miR-93–3p		oncogenic
		miR-141–3p		oncogenic
STAT3
		miR-21–5p		oncogenic
		miR-181–5p		tumor suppressor, targeting KRAS

3.6. Cross-talk between signaling hubs

There is numerous cross-talks between the signaling hubs in MWCNT treated cells. Figure 4 shows the cross-talk in 55-OH-MWCNT 30 μg/ml treated cells. MYC is at the center of the interaction since it cross-talks with most of the other signaling hubs (all except GRB2). Some of the hubs crosstalk directly, some through a gene/protein. For example, AR and TP63 both regulate Transforming growth factor beta 2 (TGFB2), Cyclin dependent kinase inhibitor 1A (CDKN1A) interacts with MYC, STAT3, BRCA1, AR, TP53 and TP63, and CD44 is regulated by GRB2 and TP53. So, altering the expression of one gene/microRNA will affect many genes/microRNA that are regulated by them or that they regulate.

Figure 4.

Cross-talk between signaling hubs in 55-OH-MWCNT 30 μg/ml treated cells.

3.7. Inflammation related genes/microRNA

To further examine the alterations in inflammation by these three MWCNTs, we listed some of the genes and microRNA expression levels in Table 6. It appeared that 55-OH-MWCNT caused more gene/microRNA changes and larger changes, 66-COOH-MWCNT is next and 33-MWCNT is the least. This might be an indication that 55-OH-MWCNT caused more inflammation, and 33-MWCNT the least.

Table 6.

Expression levels of inflammation-related genes/miRNAs

expression levels	33-MWCNT	55-OH-MWCNT	66-COOH-MWCNT
IL-6	−1.73	−1.85	−1.76
CXCL8	−2.24	3.82	−2.37
IL1R1	na	1.39	1.39
TNF	na	−6.38	na
miR-146a-5p	−2.12	−5.00	−3.50
miR-155–5p	−1.59	2.25	−2.00

Note. Expression levels are given as treated/control gene (miRNA) levels.

Positive numbers mean up-regulation; negative numbers mean down-regulation.

“na” means not significant.

3.8. Discussion

We studied three different kinds of MWCNTs, one parent compound (33-MWNCT), two functionalized MWCNTs (55-OH-MWCNT and 66-COOH-MWCNT). At the concentration we used for RNA expression (0.3 to 30 μg/ml), very little cytotoxicity was observed, all the treated samples had over 80% survival (Figure 1). For the ROS/RNS production, at 30 μg/ml, all three MWCNTs showed over 2-fold increase of ROS/RNS levels.

For the canonical pathways altered by these three MWCNTs (Table 3), there are similarities and differences. Similarities: firstly, many stress response pathways (red lettering) such as mitochondrial dysfunction, oxidative phosphorylation and protein ubiquitination pathway and many more were altered. There were 28, 23, and 21 stress response pathways altered by 33-, 55-OH-, and 66-COOH-MWCNT respectively. Secondly, many of the metabolic pathways (dark green lettering) were altered by all 3 MWCNTs in treated cells. There were 29, 17 and 9 metabolic pathways altered in 33-, 55-OH and 66-COOH- MWCNT treated cells, respectively. Cholesterol biosynthesis pathways are major pathways altered among the metabolic pathways.

There are also differences in the signaling pathways altered between these 3 MWCNTs presented in Table 3. Firstly, NRF2 mediated oxidative stress response pathway was only present in 33-MWCNT at 30 and 10 μg/ml treated cells, ranked 7 and 2 respectively. Secondly, base excision repair (BER) pathway was only present in the 55-OH-MWCNT, nucleotide excision repair pathway was only present in 66-COOH-MWCNT treated cells. Thirdly, there was no protein synthesis-related pathways (dark blue lettering) or cancer-related pathways (black lettering) altered in 33-MWCNT treated cells, but they were present in the two functionalized MWCNTs treated cells, and the protein synthesis-related pathways were ranked very high (low in numbers, low in p-value) in both 55-OH- and 66-COOH-MWCNTs treated cells. From Table 3, the two functionalized MWCNTs are more similar to each other than to the parent compound.

When the signaling pathways were ranked according to their p-values (Table 4), we observed the following:

1. Stress response pathways: (A), the NRF2 mediated oxidative stress response pathways was affected in almost all doses of the MWCNT, except 33-MWCNT at 1 μg/ml treated cells. Only in 33-MWCNT treated cells, were they ranked in the top 12. In the two functionalized MWCNTs, they were ranked much lower (higher in number, lower in probability), and therefore, did not show up in Table 3. The presence of the NRF2 mediated signaling pathway in all 3 MWCNT treated cells correlated with our results showing ROS/RNS induction after MWCNT treatment. This also correlated well with literature that MWCNTs induced oxidative stress in treated cells [27]. (B), many altered signaling pathways are direct results of oxidative stress, these include mitochondrial dysfunction, oxidative phosphorylation, and hypoxia inducible factor one alpha (HIF1α) signaling. These signaling pathways were all altered in all three MWCNTs treated cells. Once again, this correlated well with our ROS/RON production data. (C), two DNA damage repair pathways, NER and BER, were altered by all three MWCNTs, but to a greater extent by the functionalized nanotubes, indicating that the cells treated with the functionalized MWCNTs may have more DNA damage and/or less DNA repair capacities. This is consistent with literature reporting that the functionalized MWCNTs caused more DNA damage [28]. Our DNA damage experiments did not detect increased DNA damage. One possible explanation is the gene expression analysis is more sensitive than the methodology (protein analysis or the florescence methods) we used to detect the DNA damage. Another possible explanation is the kind of DNA damage assays we used. We only looked at apurinic/apyrimidinic (AP) sites and 8-hyrdoxy-2’-deoxyguanosine (8-OHdG) formation. There are many other types of DNA damages like base loss through hydrolysis, alkylation of the base, bulky adduct formation, DNA cross link and DNA strand breaks. The kind of DNA damage that may have been caused by MWCNTs might not be the ones we assayed for.

2. Some of the signaling pathways secondary to the oxidative stress production were also altered. (A), unfolded protein response (UPR), protein ubiquitination pathways (PUP), ferroptosis, sirtuin signaling pathways and inflammation-related pathways, all these signaling pathways were indirectly affected by oxidative stress, and they were altered in all three MWCNTs treated cells in most of the concentrations examined, indicating ROS production in these treated cells. Through these signaling pathways, many different signaling pathways and biochemical pathways would be altered, such as lipid metabolism, immunity, inflammation, DNA damage response, cell cycle controls, mitochondrial biogenesis, protein synthesis, cell cycle controls, cell death and many more [29–36]. (B), protein synthesis-, cell cycle- and cancer- related pathways, these pathways were affected much more by the two functionalized MWCNTs than the parent MWCNT; especially in the protein synthesis-related pathways, the two functionalized MWCNT caused more significant (rank low in numbers, high in p-value) alterations.

3. Cholesterol biosynthesis pathways: Cholesterol biosynthesis takes place both in the cytosol and the endoplasmic reticulum (ER). The rate limiting enzyme in the cholesterol biosynthesis is 3-hydroxy-3-methyl-glutaryl-coenzyme A reductase (HMG-CoA reductase), which catalyze the reduction of HMG-CoA to mevalonate resides in the ER membrane. And because the unfolded protein response and the ER stress pathway were affected in many of the MWCNTs treated cells, it is not surprising to see many of the cholesterol biosynthesis pathways altered in these cells.

In summary, cholesterol biosynthesis pathways and some of the oxidative stress related pathways, directly or indirectly, like NRF2 mediated stress response, mitochondrial dysfunction and oxidative phosphorylation, UPR and PUP, and inflammation-, cell cycle-, and cancer-related pathways were altered in all three MWCNTs treated cells. This correlated well with our data showing all three MWCNTs increased ROS/RNS in treated BEAS-2B cells. However, our data also showed that the two functionalized MWCNTs induced more alterations in DNA repair pathways, cancer-, protein synthesis, and cell cycle-related pathways. This may be an indication that the functionalized MWCNTs are more genotoxic than the parent compound. The mechanism of this difference awaits more studies.

From Figure 3 a to c we can see the signaling hubs in each of the MWCNT treated cells. The hub that is centered on the MYC gene is present in all three MWCNTs treated cells. MYC is a proto-oncogene, homologous to the v-MYC carried by the Avian Myelocytomatosis virus. MYC protein can directly activate CCND1, and thus induces cell proliferation [37]. MYC gene also cross talks with many genes and affects many signaling pathways. In 33-, and 66-COOH-MWCNT treated cells, both MYC and CCND1 hubs are present, meaning MYC could induce cell proliferation through activating CCND1 in these cells. The CCND1 hub was not present in 55-OH-MWCNT treated cells. However, there are other hubs in 55-OH-MWCNT treated cells that can interact with MYC and cause cell proliferation, such as STAT3 and AGO2 [38, 39]. Even the hubs that regulate inflammation such as AR and GRB2, may lead to proliferation through inflammation [40, 41].

The hubs that are present in both 55-OH-, and 66-COOH-MWCNT treated cells, but not in 33-MWCNT treated cells are GRB2, AR, TP63 and AGO2. GRB2 is an adaptor protein, that once bound by growth factors, will activate RAS and its down-stream Mitogen activated protein kinase (MAPK) cascade and induce cell proliferation [40]. TP63 is one of the transcription factors that transactivates TP53 target genes [42]. In addition, TP63 has been shown to regulate a diverse set of microRNAs including the miR-17 family (miR-17, miR-20b and miR-106a) which were altered in both 55-OH-, and 66-COOH-MWCNT treated cells (data not shown) [43]. The microRNA-17 family is an oncogenic microRNA family. Therefore, a change in TP63 expression would result in a change in the proliferation of the cells. TP63 also regulates miR-34a (altered in both 55-OH- and 66-COOH-MWCNT treated cells) which in turn regulates TP53 [44].

AGO2 is a component of RNA-induced silencing complex (RISC) [45], so, altering AGO2 expression will alter many signaling pathways by changing microRNAs’ interactions with their target mRNAs. In addition to involving in the RISC complex, AGO2 also physically interacts with RAS protein and promotes its downstream signaling of cell proliferation [39]. Androgen is the dominant male sex hormone; it drives the development and maintenance of the male characteristics by binding to the androgen receptor. However, androgen is systematically distributed throughout the whole organism. Recent research has shown that androgen also regulates the immunity in the organism [46]. In the lungs, many immune cells express ARs, and are responsive to androgen. Androgen suppresses inflammation and allergic reaction in the lung [47].

AR and GRB2 affect inflammation response and were not present in 33-MWCNT treated cells. This indicated that 55-OH-, and 66-COOH-MWCNT altered inflammation pathways more than the parent 33-MWCNT. This correlated with more inflammation related signaling pathways altered in the canonical pathways resulted from IPA analysis of the mRNA transcription analysis (Table 4). In short, the two functionalized MWCNT altered more signaling hubs that regulate inflammation and cell proliferation than the non-functionalized parent MWCNT.

Signaling hubs that were only present in 55-OH-MWCNT treated cells were TP53, STAT3 and BRCA1. STAT3 is often constitutively expressed in lung cancer; it promotes cell proliferation, angiogenesis, immune evasion and ani-apoptosis [48]. TP53 is a well-known tumor suppressor, known as the guardian of the genome [49], involved in DNA damage repair, G1, G2/M cell cycle arrest, apoptosis, senescence and more [50]. BRCA1, another tumor-suppressor, known as the caretaker of the genome [51], is also involved in DNA damage repair, transcriptional regulation and ubiquitination regulation. There is cross-talk between TP53 and BRCA1. Alteration of these 2 hubs might be an indication that there was more DNA damage in the 55-OH-MWCNT treated cells, which correlates with DNA damage repair pathways analysis of the mRNA DEGs (Table 3). However, our DNA damage assay did not get positive results. Again, we only looked at AP-sites and formation of 8-OHdG. It could be the 55-OH-MWCNT caused DNA damages other than these 2 kinds of DNA damage that we measured. In short, 55-OH-MWCNT affected signaling hubs that are important for genome integrity and cell proliferation. This might indicate that 55-OH-MWCNTis more genotoxic in the treated cells.

There are many cross-talks between the gene hubs in the microRNA target filter analysis results. In addition to genes, there are microRNA in these hubs. One microRNA can interact with multiple genes. For example, miR-200 interacts with MYC gene, but also with AGO2. MiR-146a-5p interacts with CCND1 and also with AR. Therefore, a single change in one miR/gene level may results in a cascade of changes. In BEAS-2B cells, the signaling pathways are interactive and inter-related and are connected in multiple very complicated ways. These facts show the complexity of the signaling pathway webs that are changed by the MWCNTs that we used.

Overall, our in vitro data in MWCNT treated human lung epithelial cells (BEAS-2B) correlated well with the already published in vivo effects of MWCNTs which includes sustained inflammation, fibrosis, DNA damage, and lung cancer following long-term inhalation exposures [4–6]. Our signaling pathways that directly relate to the in vivo results were: NRF2-mediated oxidative stress pathway, inflammation pathways (TGFβ signaling, TNFR1signaling), DNA damage repair pathways (BER and NER), and cancer related pathways. Signaling pathways that indirectly relate to the in vivo effects of MWCNTs include mitochondrial dysfunction, TP53 signaling, and protein ubiquitination pathway. Signaling hubs from microRNA target analysis that were related to the in vivo effects include inflammation related (AR and GRB2) and cancer-related pathways (MYC, CCND1, TP53, TP63, AGO2, and STAT3).

4. CONCLUSIONS

We studied MWCNTs (20–30 nm in outer diameter) with or without functionalization. The functionalization was either hydroxylation (-OH) or carboxylation (-COOH). We treated BEAS-2B cells with all three MWCNTs and studied the cytotoxicity, DNA damage, ROS/RNS production and gene expression. We did not detect DNA damage with our assays for 8-oxo-dG or AP sites but did observe ROS/RNS production in all three MWCNTs treated cells. For the gene expression analysis (both mRNA and microRNA), 55-OH-MWCNT caused more changes in DEGs for both mRNA and miRNA, and 33-MWCNT caused the least changes in DEGs. From mRNA expression analysis, we detected many stress-related pathways. Cholesterol biosynthesis pathway was altered by all three MWCNTs. Some signaling pathways that were secondary to oxidative stress such as unfolded protein response, protein ubiquitin pathway, sirtuin pathway, and ferroptosis pathway was altered by all three MWCNTs to similar degrees. Protein synthesis related pathways, cell cycle pathways and cancer related pathways were larger in response and rank higher in 55-OH- and 66-COOH-MWCNTs treated cells.

From microRNA target filter analysis, there were altered signaling hubs. MYC is the only signaling hub affected by all 3 MWCNTs in treated cells. MYC regulates cell proliferation, therefore, changes in MYC signaling indicated possible changes in the proliferation rate of all three MWCNT treated cells. Signaling hubs that are common to 55- OH-and 66-COOH-MWCNTs are GRB2, AR, TP63 and AGO2. These hubs together with the microRNA they interact with regulate cell proliferation (TP63 and AGO2) and inflammation (AR and GRB2). These signaling hubs are not seen in 33-MWCNT, indicating 33-MWCNT might have caused less alteration in inflammation. The signaling hubs that only present in 55-OH-MWCNT are TP53, STAT3 and BRCA1. TP53 and BRCA1 both play important roles in keeping the genome integrity. Alteration in these two genes may indicate that 55-OH-MWCNT is more genotoxic than the other two MWCNTs. There are reports that functionalized MWCNTs are more genotoxic than the unfunctionalized MWCNT (Zhou et al., 2017), but there was no data on comparison between -OH and -COOH functionalized MWCNTs with the same outer diameter. From the gene/microRNA expression levels of inflammation related genes, 55-OH-MWCNT caused most inflammation, and the 33-MWCNT, the least. This correlated well with most gene/microRNA changes (DEGs) for 55-OH-MWCNT and least for 33-MWCNT (at 30 ug/ml).

In conclusion, our study indicated that functionalized MWCNT are more genotoxic (DNA damage) than the unfunctionalized MWCNT using the DNA repair pathways (BER and NER) as surrogate indicators. Furthermore, we showed that -OH modified MWCNT are more potent in causing changes in DNA repair pathways than the -COOH modified MWCNT. This correlated well with the in vivo data where functionalized MWCNT are more potent in causing lung tumorigenesis. We also demonstrated that functionalized MWCNTs caused more and larger changes in expression of the inflammation related genes/miRNAs. This may translate into more inflammation in functionalized MWCNTs treated cells. In addition to changed expression of genes and miRNAs, we also demonstrated there were extensive cross talks between these signaling hubs, demonstrating the effects of these MWCNT are far reached and caused more long term damages. We believe this is the first time the complexity of the networks and cross-talks has been demonstrated in detail in MWCNT treated cells. Finally, we demonstrated that the in vitro signaling pathways or hubs correlated well with the published, in vivo, pathological effects induced by MWCNTs, such as inflammation, fibrosis of the lung and lung cancer.