Transcriptome profiling by RNA sequencing reveals novel targets of Gemini nano curcumin on p53-mutant HT-29 colorectal cancer cells

Abstract

Introduction

Colorectal cancer is considered as an aggressive tumor with high mortality in the world. It has been shown that Gemini Nano-Curcumin (Gemini-Cur) affects viability of colorectal cancer cells. Nevertheless, the cellular and molecular mechanisms underlying its toxicity are debatable. Here, we planned to untangle the potential novel targets for nanocurcumin on p53-mutant HT-29 cancer cells by employing RNA sequencing.

Methods

After cultivation, HT-29 cell were incubated with appropriate doses of Gemini-Cur for 24 h. Total RNA was extracted, cDNA library was constructed and sequenced. The DESeq2 tool were employed to normalize reads and detect differentially expressed genes (DEGs). The enrichR tool was employed to do Gene Ontology (GO) & identify biological processes (BP), cellular components (CC) and molecular functions (MF) that are impacted in the condition. The PPI network was constructed of 1200 DEGs using STRING, visualized by Cytoacape and analyzed with MCODE.

Results

After DEGs screening, 1309 genes between untreated and treated cancer cells were obtained including 479 upregulated with P-value < 0.05 & log2 FC > 1) as well as 63 downregulated genes with P-value < 0.05 and log2 FC < −1). Then, DEGs were assigned to 207 GO terms including numerous cellular pathways such as endoplasmic reticulum (ER)-related processes. Finally, 542 up and downregulated genes were mapped to 67 reactome pathways. The pathway analysis illustrated that Gemini-Cur modulates numerous pathways including ER stress response- and transporter-related genes. Using MCODE, Three modules were significantly identified with scores ≥ 1 and nodes ≥ 1.

Conclusion

Our data reveal that nanocurcumin might affects HT-29 colorectal cancer cells through modulating different pathways such as endoplasmic reticulum response, and transporter-related genes. More studies necessitate to unravel the molecular mechanisms and proteins involved in these pathways that could be considered as novel therapeutic targets in colorectal cancer.

1. Introduction

Colorectal cancer (CRC) is considered as the second leading cause of malignant-based deaths after lung cancer in the world1, 2. Although the findings show that the death rate of CRC is decreasing worldwide, its early-onset incidence has significantly increased. The early diagnosis of colon cancer is usually explained as the occurrence of tumor nodules in patients under 50 years of age³. Performing genetic tests and examining family history can be very helpful in preventive measures for people with a hereditary tendency to develop neoplasms1, 4, 5. Regardless of recent appropriate tools being applied for early detection and treatment of CRC, developing new treatment strategies with fewer side effects is needed to relieve symptoms and lengthen the mean survival of patients⁶.

Gemini curcumin, a type of phytochemical, has several advantages due to its enhanced cellular absorption and bioavailability and toxicity. Gemini surfactants represent a category of nanoscale substances characterized by two identical structures connected by a spacer. These compounds exhibit remarkable efficacy in the transportation of genes and pharmaceuticals. Gemini-Cur can induce programmed cellular death in tumors by influencing cell cycle program and promoting gene expression⁷. In a study, Gemini-Cur was used as a stable nano-formulation on HT-29 cells⁸. It was shown that this nano-formulation significantly increases the cellular uptake, stability and cytotoxicity of curcumin⁸.

On the other hand, bioinformatics databases such as the Gene Ontology (GO) database (https://geneontology.org), can be beneficial as a comprehensive source of the function of genes and theirs products, including proteins and non-coding RNAs⁹. Through analytical and systematic studies, hub genes related to CRC were identified to understand the networks and paths of disease progression¹⁰. Among these, 33 hub genes (CXCL3, MYC, CCNB1, CXCR1, BGN, SPP1, COL1A2, UBE2C, LPAR1, CCNA2, GUCA2A, CXCL2, NMU, PPBP, THBS1, CXCL11, TIMP1, CDK1, CCMP1, TOP2A, CXCL8, AURKA, BMP2, SST, COL1A1, CXCL5, PYY, MMP3, CND1, GCG, CXCL12, PAICS and THBS2) with clinical significance in colorectal cancer-related pathways were introduced10, 11. Further analysis also revealed that the key upstream regulators including SMAD3, RELA, SOX2, STAT3, FOXM1, and NF-κB are associated with the central (core) genes¹⁰. These genes may promote the screening and identification of essential biomarkers and targeted treatment of colorectal cancer¹¹.

RNA sequencing (RNA-Seq) is employed to study the whole RNAs of a biological samples. Compared to other technologies, this tool provides a deep coverage of the variable nature of the gene functions¹². Additionally, the generated data assist the discovery of new transcripts and variantsas well as the determination of allele-specific expressions¹². Recent developments in RNA analyses, have enabled scientists to further uncover the functional complexity of transcriptome¹². Therefore, we employed RNA sequencing to partly unravel the molecular targets of Gemini-Cur on p-53 mutant colon cancer and, to identify the most important modulated genes and pathways in HT-29 cells.

2. Materials and Methods

2.1. In vitro

2.1.1. Reagents and cell culture

We used frozen HT-29 cells available at the lab and DMEM medium was employed as culture solution that was accompanied with 10 % (v/v) FBS (fetal bovine serum) and 1 % (v/v) solution of penicillin/streptomycin. Curcumin and mPEG urethane gemini surfactant nanoparticles were obtained as a gift from Institute for Color Science and Technology in Tehran, Iran.

2.1.2. Preparation of Gemini curcumin formulation

To prepare Gemini-Cur nanoparticles, we used nanoprecipitation method previously set up at our lab⁶. Briefly, both compound in appropriate concentrations were added to three milliliter of methanol and in a gently stirring condition, the solution was diluted twice with sterile PBS. After evaporation of methanol in rotary evaporator, the mixture was filtered through a 0.22 µM syringe to eliminate any contaminations and kept at 4 ◦C until use.

2.1.3. Nano-compound treatments

The human colorectal carcinoma cells HT-29 was seeded on 6-well plates in a high-content glucose DMEM (DMEM/HG; Gibco) culture medium. After 24 h, Gemini-Cur was added to the plates in triplicate and incubated for another 24 h. All the treatments were done in triplicate.

2.1.4. Preparation of RNA for sequencing

The TRIzol reagent was employed to obtain pure RNA. After evaluation of the quality and quantity (NanoDropTM, ThermoFisher, USA), all RNAs were incubated with RNase-free DNase I (Takara, Japan). Finally, RNAs were dried in GeneTegra-RNA tubes (GenTegra Co., Korea), by freeze dryer (Sartorius Co., Germany) and sent to Macrogen company for sequencing (Macrogen Co., South Korea).

2.1.5. Library construction & RNA sequencing

Library construction was done by Macrogen Company. Briefly, One µg of RNA obtained from treated cells was employed to synthesis cDNA libraries using TruSeq RNA Sample Prep Kit v2 (Illumina, Inc., USA). After cDNA synthesis, they were attached to barcoded adapters and amplified by PCR. Also, RNA spike-in control Mix 1 (Life Technologies Corporation, USA) was added to the tubes13, 14, 15. The amplified cDNA fragments were sequenced using a HiSeq 2000 sequencing system (Illumina, Inc., USA). All the sequencing records were stored in raw FASTQ format for further analysis.

2.2. In silico analyses

2.2.1. Quality check of RNA sequencing reads, mapping and read Annotation

Raw data were processed and analyzed using Ubuntu 22.04 and R/Bioconductor as an online available tool (Table 1, Table 2). The paired-end reads were checked and trimmed with MultiQC (https://github.com/ewels/ MultiQC, 25 January 2022) and Trimmomatic (http://www.usadellab.org/cms/?page= trimmomatic, 25 January 2022), respectively. The genome reference GRCh37 (hg19) were employed to map clean reads by STAR (https://github.com/alexdobin/STAR/ releases, 10 January 2022). Finally, transcripts were counted with Htseq-count (https://htseq.readthedocs. io/en/release_0.11.1/count.html, 10 January 2022).

Table 1.

The list of the samples.

R	Samples	Library Layout	Number	Experimental Design (Tumor/Normal)
1	UT1. fastq	PAIRED	1	Untreated
2	UT2. fastq	PAIRED	1	Untreated
3	T1. fastq	PAIRED	1	Treated
4	T2. fastq	PAIRED	1	Treated

Table 2.

Data and reference specifications.

R	Data	Reference Specifications
1	Organism	Human
2	Genome	GRCh38
3	Annotations	(gtf)

2.2.2. Read counts normalization, differentially expression Analysis, and network construction

The Bioconductor workflow (DESeq2) was used to normalize the counts and determine differentially expressed genes¹⁶. Then, the upregulated (p value < 0.05, log2 FC > 1), and downregulated genes (p value < 0.05, log2 FC < −1) were obtained. Accordingly, STRING was employed to construct PPI network. Cytoscape tool (version 3.6) visualized the interactions in the PPI network with cut-off value of BC > 0 and K > 8. Then, MCODE tool was employed to detected modules with score ≥ 1 and nodes ≥ 1. Then, DEGs were mapped to related pathways in KEGG¹⁷, and the annotated network of significant KEGG pathways was visualized in Cytoscape18, 19, 20, 21.

2.2.3. Gene ontology studies

Gene ontology (GO) analysis helps to understand the functional roles of DEGs by mapping them to specific GO terms. This allows researchers to identify which cellular components (CC), molecular functions (MF), or biological processes (BP) are most affected by the gene expression changesGene Ontology enrichment and pathway analysis was performed on the differentially expressed genes using Enrichr/Bioconductor package. GO annotations were obtained to clarify which pathways and functional biological categories the DEGs are enriched²².

3. Results

3.1. Quality Assessment and statistics of raw data

After quality check for the reads with Fastqc and Trimmomatic, all the cleaned reads were retrieved for further analysis. The proportion of bases with Q20 demonstrated the proper quality of reads from RNA sequencing. The replicates of untreated and treated samples were assigned for RNA sequencing. Totally, more than 40 million reads were recorded in each sample (Fig. 1). The reads were mapped to the genome reference GRCh38, and the efficiency of 80.50 % were accounted. After elimination of adaptors and quality assessments, over 5 million reads were assembled for each sample (Fig. 1D).

Fig. 1.

Distribution of read counts. Distribution of raw read counts (log2 (counts + 1)) are shown as density and Bar plots (A & B). The normalized counts are shown as density plot (C). The line plot confirms over 5 million reads for each sample (D).

3.2. Identification of genes with differentially expressed patterns

The Htseq-count and DESeq2 tools were performed for counting transcripts and detection of genes with differentially expressed pattern in treated cells versus untreated group. In total, 1309 DEGs including 479 up-regulated (p value < 0.05, log2 FC > 1) and 63 down-regulated (p value < 0.05, log2 FC <  − 1) genes were determined in this study (Supplementary Material). To better understanding the modulation of expression, top genes with up and downregulations were considered in color heatmaps. The color ranges from green (down) to purple (up) for non-normalized (Fig. 2A) and normalized data (Fig. 2B). The distribution of gene groups with different values are shown in volcano plot (Fig. 2C).

Fig. 2.

Heatmap plot for top modulated genes associated with non-normalized (A) and normalized counts (B). Volcano plot of all genes modulated in treated cells. NS: Non significant.

Table 3, Table 4 show the top 10 genes with highly modulated expressions in treated cells.

Table 3.

The list of top 10 upregulated DEGs (p value < 0.05, log2 FC > 1).

Gene Symbol	baseMean	Log2FoldChange	lfcSE	Stat	P-value	Padj
ATP13A4	195.9021	4.980841	0.94088	5.29381	1.20E-07	0.001842
FOSB	1128.832	3.399522	0.684332	4.967648	6.78E-07	0.003907
CYP4F11	193.182	3.580583	0.724119	4.944745	7.62E-07	0.003907
AKR1C2	886.5594	2.344592	0.480731	4.877137	1.08E-06	0.004137
AKR1C1	1647.562	2.059895	0.465552	4.424628	9.66E-06	0.024756
FTX	546.2693	3.354606	0.790279	4.244839	2.19E-05	0.042041
MT1G	71.29674	3.790338	0.91703	4.133276	3.58E-05	0.054985
SLC7A11	4543.659	2.550962	0.626592	4.071167	4.68E-05	0.065383
DYNC2H1	879.8413	2.810885	0.707901	3.970731	7.17E-05	0.077682
HACD4	23.25406	8.040045	2.035359	3.950185	7.81E-05	0.077682

Table 4.

Top 10 downregulated DEGs with p value < 0.05 & log2 FC <  − 1.

Gene symbol	baseMean	Log2FoldChange	lfcSE	Stat	P-value	Padj
HSPA6	837.8478	−2.57765	0.549342	−4.69225	2.70E-06	0.008309
LGALS9C	27.04325	−4.06867	1.248232	−3.25954	0.001116	0.241997
PADI2	388.9813	−1.81928	0.57882	−3.14308	0.001672	0.2856
ALDH1A3	358.4845	−1.94937	0.633494	−3.07717	0.00209	0.336182
AKAP12	224.0475	−1.9651	0.643404	−3.05422	0.002256	0.354014
DIRAS2	40.10447	−3.13951	1.057168	−2.96974	0.002981	0.380594
SLAMF9	39.23086	−3.5505	1.279141	−2.77569	0.005508	0.476126
HSPA1A	1457.486	−1.33674	0.482678	−2.76942	0.005616	0.478645
ARSI	13.77364	−5.6986	2.076701	−2.74406	0.006068	0.493893
NDUFA4L2	3876.594	−1.64373	0.607141	−2.70732	0.006783	0.513815

3.3. Gene ontology and pathway analysis

The Enrichr web tools were employed to enrich DEGs in related pathways and ontology categories. After analysis, the DEGs were significantly assigned to 207 GO categories including 169 (BP), 11 (CC) and 27 (MF) as shown in Table 5. Accordingly, about 542 modulated genes with p value < 0.05; log2 FC > 1 for upregulated genes, and with p value < 0.05; log2 FC < −1 for downregulated genes were mapped to 67 Reactome pathways. Table 6 illustrates the top 10 enriched DEGs in the reactome pathway database.

Table 5.

Gene ontology data for DEGs. CC: Cellular Component, MF: Molecular Function, and BP: Biological Process.

Term and GOID	Source	P value	Genes
Response to endoplasmic reticulum stress (GO:0034976)	BP	2.43E-06	NIBAN1; FICD; HSPA5; UBA5; RASGRF2; SEL1L; EIF2AK3; SYVN1; HSP90B1; UFL1; ERLEC1; SELENOS; CREB3L2; HYOU1
Co-translational protein targeting to membrane (GO:0006613)	BP	1.51E-05	SEC61A1; TRAM1; SRP72; SSR1; SEC63
Endoplasmic reticulum membrane (GO:0005789)	CC	1.56E-05	ERO1B; TRAM1; MCTP1; ABCB6; FICD; SAR1B; UBA5; SEL1L; PIGZ; MIA2; MIA3; MYRFL; CYP19A1; SELENOI; MTDH; HSP90B1; CLGN; CDC42; SEC61A1; UFL1; ADGRG6; SSR1; GPAT3; UGT1A6; ABCA1; CERS3; SLC30A7; HSPA5; PNPLA8; EDEM1; VPS13C; CYP4F3; EIF2AK3; SYVN1; VPS13A; CYP4F11; SPCS3; VAMP7; SELENOS; SCD; ATP13A4; SRPRB; SEC24D
Hrd1p ubiquitin ligase ERAD-L complex (GO:0000839)	CC	0.006893072	SEL1L; SYVN1
alditol: NADP + 1-oxidoreductase activity (GO:0004032)	MF	0.0020385	AKR1B10; AKR1C1; AKR1C2
Bitter taste receptor activity (GO:0033038)	MF	0.004900586	TAS2R43; TAS2R14; TAS2R19; TAS2R4

Table 6.

The list of genes involved in the reactome pathway database.

Pathways	P value	Genes
Unfolded protein response (UPR) R-HSA-381119	1.42E-04	HSPA5; FKBP14; EDEM1; CREB3L2; EIF2AK3; SYVN1; SSR1; SRPRB; HYOU1; HSP90B1
ER quality control compartment (ERQC) R-HSA-901032	2.00E-04	EDEM3; EDEM1; SEL1L; SYVN1; UGGT1
IRE1 alpha activates chaperones R-HSA-381070	2.84E-04	HSPA5; FKBP14; EDEM1; SYVN1; SSR1; SRPRB; HYOU1
Glucuronidation R-HSA-156588	4.78E-04	UGDH; UGT1A1; UGT1A3; UGT1A8; UGT1A6
Calnexin/calreticulin cycle R-HSA-901042	5.79E-04	EDEM3; EDEM1; SEL1L; SYVN1; UGGT1
Aspirin ADME R-HSA-9749641	0.00111256	ABCC2; SLC16A1; UGT1A1; UGT1A3; UGT1A8; UGT1A6
NR1H2 And NR1H3-mediated signaling R-HSA-9024446	0.001410707	ABCA1; ARL4C; SCD; UGT1A3; ABCG1; TNRC6A
XBP1(S) activates chaperone genes R-HSA-381038	0.001410707	FKBP14; EDEM1; SYVN1; SSR1; SRPRB; HYOU1
Regulation of HSF1-mediated heat shock response R-HSA-3371453	0.001532284	NUP214; RANBP2; HSPA5; HSPA6; ATM; HSPA13; HSPA1B; HSPA1A
Transport of small molecules R-HSA-382551	0.001671553	ABCB6; SAR1B; SEL1L; AQP7; SLC5A1; SLC7A11; ATP2C2; ABCA12; CLCN3; SLC7A1; SLC7A2; BEST1; SLC9A1; ERLEC1; CA1; SLC22A15; LCN15; LRRC8D; MFSD4B; SKP1; ABCA1; SLC30A7; SLC38A1; SLC16A1; ABCC2; ATP8B2; ATP2B1; SLC6A9; TRPV6; PSME4; ATP13A4; STEAP2; ABCG1

3.4. Protein-protein interaction (PPI) network, module identification, and functional annotation analysis

To further study of the potential interactions between DEGs, the PPI network was constructed at protein level by using STRING tool. The network detected interactions between 180 nodes based on a confidence score of 0.007 (p value: 1.0e-164). Then, the interaction pairs in the network were illustrated by the network analyzer plugin and Cytoscape with a cut-off value for BC > 0 and K > 8. The MCODE tool detected 10 modules that have been summarized in Table 7. About six Modules were significantly obtained with scores ≥ 1 and nodes ≥ 1 (Fig. 4), of which 3 of the most important ones are in detail listed in Table 8.

Table 7.

The details of topological parameters for PPI network. The hub nodes in the network are based on the cut-off values of BC > 0 and K > 8.

Gene Symbol	Degree (K)	Betweenness (BC)	Closeness Centrality (CC)
HSPA5	41	0.144966049	0.399271845
ESR1	40	0.151038248	0.396385542
HSP90B1	34	0.066548091	0.374715262
PPARG	28	0.123553089	0.373439274
EGF	27	0.069025687	0.361936194
SEC63	24	0.034627382	0.320662768
CDC42	23	0.112812338	0.348886532
NFKBIA	22	0.037958139	0.360745614
SEC61A1	22	0.024531669	0.328343313
ATM	21	0.063001644	0.360350493

Fig. 4.

Six models were identified with MCODE tool.

Table 8.

The top 3 subnetworks of PPI networks were identified with MCODE scores ≥ 1 and nodes ≥ 1.

Cluster	Score (Density*#Nodes)	Node	Edge	Node IDs
1	8	8	28	NDUFB7, MT-ND2, MT-ND1, MT-ND4, MT-ND5, MT-CYB, NDUFB4, UQCRC1
2	6.615	14	43	UTP20, NOC3L, SRP72, RRP12, WDR36, UTP18, SEC61B, RPL18A, SEC61A1, RPL36A, SRPRB, PUM3, ESF1, UTP15
3	6.571	8	23	UGT1A3, CYP19A1, UGT1A6, UGT1A1, UGDH, UGT1A8, AKR1C1, AKR1C2

4. Discussion

Despite recent advancements in the early detection and appropriate management of colorectal cancer (CRC), it continues to be recognized as the third most significant contributor to cancer-related mortality on a global scale²³. Gemini-Cur represents an appropriate formulation of curcumin that our research group recently produced. This formulation has a noteworthy anticancer impact²⁴. Studies have documented that the compound Gemini-Cur effectively suppresses the growth of several cancer cells by promoting apoptosis17, 24, 25.

Despite numerous studies on the toxicity of curcumin on tumor cells, its molecular targets is controversial. Here for the first time, we used RNA sequencing technology to unravel the most important cellular pathways and genes that are modulated by Gemini-Cur in p53-mutant colon cancer cells. Our data confirm the toxic effect and show that Gemini-Cur modulates numerous genes that might involve in different cellular pathways. A recent study by RNA sequencing reported that curcumin derivative NC2603 modulates different pathways like migration and metastasis in breast cancer cells²⁶. Another work on transcriptome profiling, showed that curcumin induces ferroptosis-related pathways in cancer cells²⁷.

It has been shown that the upregulated genes such as ATP13A4, CYP4F11 and HCAD4 in Gemini-Cur treated cells are involved in cancer. The upregulation of HCAD4 in our study, is in accordance with a previous work reporting the inactivation of HCAD4 tumor suppressor in colon cancer²⁸. In another work, it was demonstrated that curcumin upregulates CYP4F11 in melanoma cancer cells²⁹. Subsequently, the downregulated genes reported in table 4 have been considered as modulated proteins in cancer. Shen et al., in 2022 showed that ARSI is highly expressed in head and neck carcinoma³⁰. In another study, it was demonstrated that HSPA6 is upregulated in glioblastoma³¹. Downregulation of these genes by Gemini-Cur can further confirm its complex anticancer property.

The identification of statistically overexpressed GO terms illustrated that the most of the DEGs in Gemini-Cur treated cells, are enriched in cellular stress, endoplasmic reticulum and signaling pathways. The Response to endoplasmic reticulum stress is a cellular process that plays crucial role in restoring normal ER function and hemostasis. It has been shown that, this response is activated through several pathways such as stress response³². Therefore, Gemini-Cur may partly affects its toxicity on cancer cells through modulation of genes involved in these processes. Similarly, the analysis on reactome database data confirm that ER stress pathways like unfolded protein response (UPR) and transport processes are affected by Gemini-Cur. Recent studies have demonstrated that tumor cells alter protein hemostasis and produce ER stress. Subsequently, UPR initiates downstream signaling that are crucial for tumor growth, microenvironment remodeling and drug resistance³³. These findings are in concordance with previous demonstrations that curcumin induces ER stress-mediated apoptosis through induction of reactive oxygen species (ROS) production34, 35. Recently, Wang et al reported that a curcumin derivative has the potential to induce cell death in colon cancer via mitochondrial dysfunction and ER stress³⁶. Here, HT-29 colorectal cancer cells have been employed as p53-mutant model. Therefore, it could be concluded that Gemini-Cur may recruit alternative proteins and pathways to make its toxicity on p53-mutant cells through induction of p53-independent apoptosis like ER stress. Further studies are demanded to study the molecular mechanism of this effect of curcumin on cancer cells that could be considered as novel therapeutic targets in cancer.

The construction of PPI network with STRING revealed about 180 nodes with confidence score 0.007. The nodes has been visualized with more than 2000 interactions. The genes characterized as nodes in Fig. 3 demonstrate the multi-targeted effect of Gemini-Cur in treated cancer cells. This finding is similar to a report on the modulatory effect of Gemini-Cur on different cellular genes and processes in HCT-116 colorectal cancer cells . Due to the complex structure of curcumin, it possesses broad targeting potential against numerous cellular pathways and proteins in cancer (40).

Fig. 3.

The PPI Network and Subnetworks obtained by STRING tool. The yellow color shows significant nodes in the network. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

In conclusion, transcriptome profiling of Gemini-Cur-treated HT-29 cell not only approves the multi-targeting properties of curcumin but also, demonstrates crucial role of ER-related terms and pathways. Further in vitro and in vivo studies are demanded to unravel the exact molecular mechanisms of this finding that could be considered as novel therapeutic targets in cancer.

CRediT authorship contribution statement

Hewa Jalal Azeez: Methodology, Investigation, Formal analysis. Raheleh Karimzadeh: Writing – original draft, Validation, Software, Data curation. Esmaeil Babaei: Writing – review & editing, Validation, Supervision, Project administration.

Funding

Not applicable.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article: