Introduction

For the discovery of new drugs, screening of natural compounds that target the proliferation of cancer cells is important1. For libraries with hundreds to thousands of extracts, they need to be screened with high-performance screening methods. Such methods allow the screening of many compounds at different concentrations at the same time on each target cell or the combination of compounds- with uniformity and high accuracy3, 2.

Screening of extracts on cancer cell models in 2-dimensional (2D) monolayer culture is limited because the monolayer model lacks the tumor cell characteristics of physiological tumors in the body4. Meanwhile, screening done on a cancer model in 3-dimensional (3D) culture may be better for studying drug effects since the 3D culture model is more similar to the animal models (and possibly clinical trials); the 3D model more closely reflects characteristics of tumors, such as differentiation, tumor microenvironment, and distribution of hypoxia in certain populations7, 6, 5. Many methods have been developed to create 3D cells like tumors in the body; these methods include use of U-shaped bottom well , the hanging drop, and cell growth in bio-matrix8, 6.

The method of using a U-shaped bottom well is heavily used in 3D cell model studies. However, one downside is that not all cell types can develop into 3D cell mass by this method9. For hanging drop culture, the advantage is that gravity is used to precipitate the cells together and thereby stimulate the cells to stick together into 3D spheres10. This method has a disadvantage of using very gentle manipulations and is difficult to develop if screened at high throughput automation. Meanwhile, the method of using biological substrates (like Matrigel) offers great potential for the development of 3D cell model13, 12, 11. Matrigel is usually stored in frozen form, at a concentration of 10-15 mg/mL; it is thawed at 4 °C and gelated in a temperature range of 24-37 °C for 30 minutes. Matrigel promotes the differentiation of different cell lines (. prostate, salivary gland, mammary epithelium, pancreas, Schwann cells, intestinal cells, and bone cells), of primary cell lines (. sertoli cells, blood cells, cartilage cells, epithelial cells, endometrial cells, and fallopian epithelial cells), and even tissue explants (. neural crest, immature follicles, and zygote)14.

This study used matrigel to create a 3D cell model of breast cancer for the purpose of screening natural compounds that inhibit the growth of breast cancer cell . Modeling of 2D and 3D monolayer cancer cells was carried out in parallel (simultaneously) with the same evaluation agents, including Alarma Blue. The IC (half maximal inhibitory concentration) values were compared between 2D and 3D cancer cell models to evaluate and select the extract which showed different effects in these two models.

This study used 34 natural plant extracts and two control drugs (Doxorubicin and Tiparazamine) on 4 cell lines (MCF-7, CD44CD24 MCF-7, VN9, and CD44CD24 VN9 cells).

Methods

Chemicals

In the research study, the library of the 34 extract (Table 1 ), which were coded with ‘E’ as the initial label ( E1-E34), were obtained from the Division of Medicinal Chemistry, Faculty of Chemistry, University of Science, Vietnam National University Ho Chi Minh City, Vietnam. Doxorubicin hydrochloride and tirapazamine were purchased from Sigma-Aldrich.

Results

IC values of extracts are different on MCF-7 2D and 3D models

The IC results of doxorubicin and tirapazamine showed that both 2D and 3D models were successfully established for anti-tumor activity evaluation (Table 2). The IC results of the 34 plant extracts on MCF-7breast cancer cells in 2D and 3D models are summarized in Table 3.

Extracts	IC50 values (µg/mL)		Extracts	IC50 values (µg/mL)
	2D model	3D model		2D model	3D model
E4	187.5	383.7	E25	597.4	870.8
E7	248.2	332.8	E27	165.3	242.5
E8	478.7	533.1	E28	299.7	673.3
E9	701.4	653.5	E30	1476	794
E10	310	154.7	E31	257.3	308.5
E11	342.9	198.3	E32	235.2	225.1
E12	303.5	160.4	E35	4450	615.9
E13	1779	1061	E36	2187	575.5
E14	348.6	593	E37	326.1	308.8
E15	1106	639.8	E38	368.1	692.2
E16	316.8	361.9	E39	345.6	270.6
E17	159.4	232.4	E40	70	1419
E18	112.4	230	E41	526.2	2063
E19	489.5	621.8	E42	499.7	359.6
E20	86.42	168.4	E43	306	620.4
E21	57.67	71.97	E45	155	361.9
E22	83.58	87.92	E46	135.4	387.6

Extracts	IC50 values (µg/mL)		Extracts	IC50 values (µg/mL)
	2D model	3D model		2D model	3D model
E4	66.2	360.3	E25	173.6	587.8
E7	69.42	307.7	E27	80.45	214.1
E8	103.2	937.8	E28	134.9	481.6
E9	153	887.8	E30	508.4	935.9
E10	50.48	162.2	EE31	85.04	253.5
E11	58.14	217.5	E32	56.65	223
E12	61.95	162	E35	73.95	624.6
E13	262.4	1243	E36	507	1215
E14	74.52	375.2	E37	103.5	322.7
E15	258.3	699.7	E38	229.8	980.4
E16	70.91	146.4	E39	62.65	386.6
E17	20.31	56.97	E40	274.5	1648
E18	145	39.89	E41	303.3	1257
E19	31.88	227.4	E42	102.3	293.9
E20	35.6	89.56	E43	60.97	449.6
E21	26.71	252.9	E45	32.16	295.5
E22	809.5	110.4	E46	71.98	459.6

There were 12/34 extracts which showed effects on both 2D and 3D culture models. These 12 extracts were: E4, E10, E11, E12, E17, E18, E20, E21, E22, E27, E45, and E46. However, most of the extracts predominantly had effects on the 2D model. In fact, 27 extracts on the 3D models were correlated with increased resistance by the cancer cells as compared to the resistance on the 2D models. Specifically, there were 7 extracts that had an IC values in the 3D model which were lower than in the 2D culture model. The 7 extracts were: E10, E12, E15, E30, E35, E36, and E42 (Figure 2). Thus, they are potential candidates for further use in the 3D culture model of MCF-7 breast cancer.

Cells	2D model	3D model	The extracts more sensitive on 3D than 2D
MCF-7	E20, E21, E22, E40	-	E10, E12, E15, E30, E35, E36, E42
CD44+CD24-MCF-7	E4, E7, E10, E11, E12, E14, E16, E17, E19, E20, E21, E27, E31, E32, E35, E39, E43, E45, E46	E17, E18, E20	E26, E22
VN9	-	-	E15, E18, E22, E30
CD44+CD24-VN9	E4, E7, E10, E11, E12, E14, E16, E17, E19, E20, E21, E31, E32, E35, E39, E45	E7, E21	E18, E22

The results of hit extracts on CD44CD24MCF-7 in 2D and 3D models

There were 7/34 extracts that had effects on both 2D and 3D culture models. These 7 extracts were: E7, E10, E12, E17, E18, E19, E21, and E45. However, the majority of the extracts predominantly showed effects on the 2D model (Table 5). As seen in Table 5, cells grown in the 3D model showed more resistance to the effects of the 32 extracts than the cells grown in the 2D model. In particular, there were 2 extracts which had IC values in the 3D model that were lower than the values in the 2D model; those 2 extracts were E26 and E22 (Figure 3). Therefore, they are potential candidates for further research in the 3D culture model of the MCF-7 breast cancer stem cell (CSC). Comparing with the results on the MCF-7 cell line, it was observed that the CD44CD24 sub-population of MCF-7 cells has a much higher resistance to the same extracts tested.

Extracts	IC50 values (µg/mL)		Extracts	IC50 values (µg/mL)
	2D model	3D model		2D model	3D model
E4	238.9	518.2	E25	4681	722.9
E7	345.6	297	E27	497.8	345.1
E8	1287	588.2	E28	1806	613.5
E9	916.8	535.7	E30	4976	1568
E10	712.6	270.9	E31	293.5	463.9
E11	559.2	686.4	E32	403.6	347
E12	635.9	496.2	E35	5799	3004
E13	7756	2706	E36	7437	3605
E14	531.4	638.3	E37	559.3	977.5
E15	5744	1088	E38	3158	2260
E16	377.7	987.1	E39	697.3	847.6
E17	211	431	E40	267.2	1212
E18	2055	430.6	E41	1083	871.1
E19	357.5	654	E42	2136	963.8
E20	103.8	122.6	E43	247.5	504.1
E21	304.2	146.1	E45	196.1	262
E22	2964	324.1	E46	1080	417.3

Discussion

The use of bio-matrix substrates (such as matrigel) to create 3D culture models is very convenient for drug screening. Use of a gel forming method- that contains the cells on the side of the culture well in a 96-well plate- facilitates easy manipulation without disrupting the gel structure or limiting cell growth in the form of single layer in the center of the well. This method also allows the creation of a 3D cell mass with a size of 100 μm within 5 days of culture. The drug test is conducted in 48 hours such that the entire drug testing procedure can be summarized in 7 days. In order to minimize errors when comparing 2D and 3D models, we conducted all experiments with both models in parallel. For both 2D and 3D models, the threshold of extracting effect was lower than 200 µg/mL.

A number of published studies have show that 3D breast cell culture better reflect the histological, biological, and molecular features of primary tumors than the same cells cultured using traditional 2D techniques15. In a study by Imamura , on a 3D breast cancer model, the breast cell mass was found to have the presence of a hypoxic cell population7; it is for this reason that the cell mass becomes sensitive to tiparazamine. In our study, we show that 10 extracts have the same effect as tiparazamine on breast cancer cells, and that they might be suitable candidates for hypoxia-targeted drug development (Table 5). Furthermore, in their study, Imamura and colleagues also showed that expression of Ki-67 was less in 3D breast cancer cell mass than in 2D, suggesting that the greater G0-dormant subpopulation was responsible for drug resistance in 3D culture.

Many studies have show that the breast cancer cell population with phenotype CD44CD24 possesses higher tolerability to chemotherapy, hormone therapy, and radiotherapy21, 20, 19, 18, 17, 16. Thus, for the drug screening in our study, these 4 breast cancer cell lines were suitable for our evaluations: MCF-7, CD44CD24MCF-7, VN9, CD44CD24 VN9. Morever, a promising outcome from out study is the identification of 10 extracts which have a more sensitive effect on the 3D culture model than the 2D culture model. These 10 extracts include: E10, E12, E15, E18, E22, E26, E30, E35, E36, and E42. These could be suitable candidates for the next steps towards developing drugs that target the hypoxic region in breast cancer. Therapies targeting cancer cells in areas of hypoxia and studies to discern mechanisms have garnered increase interest for cancer treatment. Hypoxia-related mechanisms such as overexpression of hypoxia-inducible factor (HIF) are also important avenues of research. Inhibiting HIF activity and changing the molecules involved in HIF offer hope for identifying molecular target to inhibit tumor growth or even completely halt growth22. HIF-1 also induces an increase in adenosine 2B receptor expression, thereby promoting the enrichment of breast cancer stem cells by activating protein kinase C-δ23.

Therefore, in the study herein, it was shown that the use of a 3D model of breast cancer cell culture for drug screening reflects a huge difference in drug resistance and drug sensitivity when compared to the 2D culture model. The matrigel 3D culture model is significant for screening compounds related to hypoxia-based therapy for breast cancer.

Conclusion

Medium-throughput screening on breast cancer cell models MCF-7, CD44CD24 MCF-7, VN9, and CD44CD24 VN9, in 2D and 3D culture, with 34 extracts showed that resistance to these extracts occurred when cancer cells were cultured in 3D. Resistance to extracts also manifested in the CD44CD24 cell populations (. CSC populations). There were 12/34 and 7/34 extracts which affected MCF-7 and CD44 CD24 MCF-7 cells, respectively. For the Vietnamese breast cancer cell line (VN9), there were 5/34 and 6/34 extracts which affected the VN9 and CD44CD24 VN9 cells, respectively. Overall, our study results indicated 10 potential candidates for future drug development targeting hypoxia in breast cancer.

Abbreviations

Dox: Doxorubicin

HIF: Hypoxia-Inducible Factor

TPZ: Tirapazamine

VN9: Vietnamse breast cancer cell line #9

Conflict of interest

The authors declare that there is no conflict of interests regarding the publication of this article.

Author contributions

Nhan Phan designed the project and carried out the experiments. Khuong Pham contributed to feasibility experiments. Mai Nguyen provided the extract. Nhan Phan analyzed the data and wrote the paper with contributions from all authors. Phuc Pham, Kiet Truong and Ngoc Phan suggested the idea, corrected the scientific matters, english wording and review all paper.

Acknowledgments

This work was supported by the Vietnam National University, Ho Chi Minh City, Vietnam, under grant number A2015-18-01.