miR‐99a‐5p modulates doxorubicin resistance via the COX‐2/ABCG2 axis in triple‐negative breast cancer: from the discovery to in vivo studies

Abbreviations

ABCG2

ATP‐binding cassette subfamily G member 2

ALT

alanine transaminase

AST

aspartate aminotransferase

ATCC

American Type Culture Collection

BC

breast cancer

BET

Brunauer−Emmett−Teller

BJH

Barret‐Joyner‐Halenda

CI

confidence interval

COX‐2

cyclooxygenase‐2

CRE

creatinine

CTAB

Cetyltrimethylammonium bromide

DAPI

4',6‐diamidino‐2‐phenylindole

DLS

dynamic light scattering

DRFS

distant‐relapse free survival

FDR

false discovery rate

FTIR

Fourier‐transform infrared

GEO

Gene Expression Omnibus

H&E

hematoxylin and eosin

H&E

hematoxylin and eosin

HA

hyaluronic acid

HR

hazard ratio

IC₅₀

half‐maximal inhibitory concentration

ICP‐MS

inductively coupled plasma mass spectrometry

METABRIC

Molecular Taxonomy of Breast Cancer International Consortium

MFI

mean fluorescence intensity

MSNs

mesoporous silica nanoparticles

myh7

Myosin Heavy Chain 7

NCBI

National Center for Biotechnology Information

OS

overall survival

PEI

polyethyleneimine

PTFE

polytetrafluoroethylene

PXRD

Powder X‐ray diffraction

qRT‐PCR

quantitative real‐time PCR

RIPA

radioimmunoprecipitation

RNU43

U43 Small Nuclear RNA

SAED

selected area electron diffraction

SD

standard deviation

SDS‐PAGE

sodium dodecyl sulfate–polyacrylamide gel electrophoresis

SEM

standard error of the mean

siRNAs

short interfering RNAs

TCGA

The Cancer Genome Atlas

TEM

transmission electron microscopy

TEOS

tetraethylorthosilicate

TGA

thermogravimetric analysis

TMAH

tetramethylammonium hydroxide solution

TNBC

triple‐negative breast cancer

URE

urea

UTR

untranslated region

Dear Editor,

Breast cancer (BC) is the most commonly diagnosed cancer and the fifth cause of cancer‐related death worldwide [1]. Despite the advances in BC targeted therapies, cytotoxic chemotherapy is still widely used [2]. However, around 20%‐30% of BC patients develop metastasis after treatment as a consequence of drug resistance [3]. In this context, microRNAs have emerged as potential therapeutic targets to overcome therapy resistance [4]. Therefore, we aimed to elucidate the molecular mechanisms underlying resistance to doxorubicin, one of the most effective chemotherapeutic agents used in BC. Methods are detailed in Supplementary Materials.

To identify miRNAs involved in doxorubicin resistance, the miRNA expression profiles of MDA‐MB‐231R cells with acquired doxorubicin resistance and their parental cell line MDA‐MB‐231 were compared (Supplementary Figure S1A). miR‐99a‐5p was found as the most differentially expressed miRNA, being downregulated in resistant cells (fold change = ‐17.83, P = 0.027) (Figure 1A).

FIGURE 1.

miR‐99a‐5p increases doxorubicin sensitivity through downregulation of COX‐2 and ABCG2. (A) Volcano plot representing array results (Affymetrix GeneChip miRNA 4.0). Red dots represent miRNAs significantly upregulated in MDA‐MB‐231 cells, blue dots represent miRNAs significantly upregulated in MDA‐MB‐231R cells, and the cross symbol represents miR‐99a‐5p. (B) MDA‐MB‐231 and MDA‐MB‐231R cells were transfected with scramble miRNA (miR‐NC) or miR‐99a‐5p (miR‐99a), treated with 0.5 μmol/L doxorubicin for 48 hours, and evaluated for apoptosis by flow cytometry after FITC‐Annexin V staining. Untreated cells were included as a control (mean ± SD). (C) Volcano plot representing array results (Affymetrix Clariom S array). Red dots represent mRNAs significantly upregulated in MDA‐MB‐231R cells, blue dots represent mRNAs significantly upregulated in MDA‐MB‐231 cells, and the cross symbol represents COX‐2. (D‐E) MDA‐MB‐231 cells were transfected with miR‐NC, miR‐99a, silencer negative control (siNC), or the COX‐2 siRNAs (siCOX‐2 #1 and siCOX‐2 #2), and evaluated for expression levels of COX‐2 and ABCG2 by Western blotting. Representative Western blotting images (D) and quantification (mean ± SD) (E). (F) Luciferase activity assays of HEK293T cells co‐transfected with miR‐NC or miR‐99a and the luciferase vector containing 3’UTR of COX‐2 or control luciferase vector (mean ± SD). (G) MDA‐MB‐231 cells were transfected with siNC, siCOX‐2 #1, or siCOX‐2 #2, treated with doxorubicin (0.5 μmol/L) for 48 hours, and evaluated for apoptosis by flow cytometry after FITC‐Annexin V staining. Untreated cells were included as a control (Mean ± SD). (H) MDA‐MB‐231 cells were transfected, and the percentage of pheophorbide A efflux was measured by flow cytometry to evaluate ABCG2 activity (mean ± SD). (I) After transfection, MDA‐MB‐231 cells were incubated with doxorubicin (5 μmol/L) for 3 hours at 37°C, and intracellular accumulation of doxorubicin was evaluated by fluorescence microscopy. Representative images are presented. (J) After intravenous injection, NP‐DOX‐PEI‐miR‐99a‐HA passively targets tumors by the enhanced permeability and retention effect, whereas HA actively targets CD44 receptor in tumor cells. After cellular uptake, PEI induces disruption of lysosomal membranes, thereby delivering miR‐99a‐5p and doxorubicin into the cytoplasm. (K) Internalization of NP‐DOX‐PEI‐miR‐99a‐HA in MDA‐MB‐231 and MDA‐MB‐231 cells pretreated with HA (5 mg/mL) for 30 minutes (MDA‐MB‐231+ HA) after 15‐minute exposure to nanoparticles (25 μg/mL). Representative confocal microscopy images are presented (blue: nuclei; red: NP‐DOX‐PEI‐miR‐99a‐HA; green: plasma membrane). (L) MDA‐MB‐231 cells were injected into the mammary fat pad of BALB/C mice. When the tumor size reached approximately 50 mm³, mice (n = 10 per group) were treated 3 times per week with 10 mg/kg of nanoparticles for one month. Tumor growth curve (mean ± SEM). (M) Representative images of immunohistochemical staining of Ki‐67 of paraffin‐embedded xenograft sections. (N) COX‐2 and ABCG2 expression in xenografts determined by qRT‐PCR (mean ± SEM). (O) myh7 expression in heart tissues determined by qRT‐PCR (mean ± SEM). *P < 0.05; ** P < 0.01; *** P < 0.001. Abbreviations: ABCG2, ATP‐binding cassette subfamily G member 2; COX‐2, cyclooxygenase‐2; HA, hyaluronic acid; myh7, Myosin Heavy Chain 7; miR‐NC, miRNA negative control; PBS, phosphate‐buffered saline; PEI, polyethyleneimine; qRT‐PCR, quantitative real‐time PCR; RISC, RNA‐induced silencing complex; SD, standard deviation; SEM, standard error of the mean; siNC, siRNA negative control; UTR, untranslated region

We previously described miR‐99a‐5p as a good diagnostic biomarker downregulated in BC tissues when compared to healthy breast tissues [5]. To determine its clinical relevance in BC, we examined the association between miR‐99a‐5p expression and distant relapse‐free survival (DRFS) in a cohort of 210 BC patients (GSE22216). miR‐99a‐5p was overexpressed in patients free of progression (n = 131) compared to patients with distant relapse (n = 79) (fold change = 1.72; P = 0.019), and its low expression was associated with shorter DRFS (P = 0.004; Supplementary Figure S1B‐C). Additionally, its association with overall survival (OS) was assessed in the Molecular Taxonomy of Breast Cancer International Consortium (METABRIC, n = 1262) and The Cancer Genome Atlas (TCGA) cohorts (n = 1062), which confirmed that low miR‐99a‐5p expression was associated with poor prognosis (Supplementary Figure S1D‐E).

Therefore, we hypothesized that miR‐99a‐5p could increase doxorubicin sensitivity. miR‐99a‐5p was described to increase sensitivity to radiation and cisplatin [6, 7], but its role in doxorubicin resistance was still unexplored. We demonstrated that miR‐99a‐5p overexpression decreased the half‐maximal inhibitory concentration (IC₅₀) of doxorubicin (Figure 1B) and increased doxorubicin‐induced apoptosis in MDA‐MB‐231 and MDA‐MB‐231R cells (Supplementary Figure S1F‐G).

To elucidate the molecular mechanism by which miR‐99a‐5p sensitizes cells to doxorubicin, we compared the gene expression profiles of MDA‐MB‐231R and MDA‐MB‐231 cells by microarray analysis. To identify potential miR‐99a‐5p targets, mRNAs significantly upregulated in MDA‐MB‐231R cells were selected (Supplementary Figure S2A) and compared with predicted miR‐99a‐5p targets from miRWalk 2.0 (Supplementary Table S1). We found cyclooxygenase‐2 (COX‐2) as a potential target with the highest differential expression (fold change = 4.94; P = 0.001) (Figure 1C), and proved its upregulation in MDA‐MB‐231R cells (Supplementary Figure S2B‐C). COX‐2 is implicated not only in inflammation but also in tumorigenesis, treatment resistance, and poor prognosis in cancer patients [8].

To clarify whether miR‐99a‐5p modulates doxorubicin sensitivity through COX‐2, we confirmed that miR‐99a‐5p overexpression inhibited COX‐2 expression and that it is a direct target of miR‐99a‐5p (Figure 1D‐F, Supplementary Figure S2D). Then, we corroborated that COX‐2 downregulation sensitized BC cells to doxorubicin, as it reduced IC₅₀ values (Supplementary Figure S2E‐F) and increased doxorubicin‐induced apoptosis (Figure 1G).

The ATP‐binding cassette subfamily G member 2 (ABCG2), a transporter involved in doxorubicin efflux, was suggested to be regulated by COX‐2 [9]. Hence, we hypothesized that ABCG2 could be a downstream target of miR‐99a‐5p through COX‐2. We found that miR‐99a‐5p downregulated the expression of ABCG2 (Figure 1D‐E, Supplementary Figure S3A) and decreased the efflux of the ABCG2‐specific substrate pheophorbide A (Figure 1H), thus confirming an inhibitory effect on ABCG2 activity. Besides, we proved that overexpression of miR‐99a‐5p increased the intracellular accumulation of doxorubicin (Figure 1I, Supplementary Figure S3B). Despite ABCG2 being listed as a predicted target of miR‐99a‐5p by miRWalk 2.0, luciferase reporter gene assay showed no direct interaction (Supplementary Figure S3C). Therefore, we proposed COX‐2 as a link between miR‐99a‐5p and ABCG2. Our results evidenced that silencing of COX‐2 downregulated ABCG2 (Figure 1D,E, Supplementary Figure S3D), reduced the pheophorbide A efflux (Figure 1H), and increased the intracellular accumulation of doxorubicin (Figure 1I, Supplementary Figure S3B). Besides, ABCG2 was upregulated in MDA‐MB‐231R cells (Supplementary Figure S3E‐F). Collectively, these data suggest that miR‐99a‐5p increased sensitivity to doxorubicin in part by modulation of ABCG2 through COX‐2.

To translate our findings, we designed mesoporous silica nanoparticles (MSNs) to deliver a combination of miR‐99a‐5p and doxorubicin to tumor cells. MSNs were loaded with doxorubicin and capped with consecutive layers of polyethyleneimine (PEI), miR‐99a‐5p, and hyaluronic acid (HA) (NP‐DOX‐PEI‐miR‐99a‐HA) (Supplementary Figure S4A). PEI favors lysosomal escape, which is necessary to ensure the miRNA delivery into the cytoplasm, and HA is a biocompatible molecule that acts as a cap, inhibiting the release of the nanoparticle's content before internalization by cells, and targets CD44, which plays an important role in BC drug resistance (Figure 1J). Empty nanoparticles (NP‐PEI‐HA), nanoparticles containing only miR‐99a‐5p (NP‐PEI‐miR‐99a‐HA), or only doxorubicin (NP‐DOX‐PEI‐HA) were evaluated as controls.

Nanoparticles were characterized by standard techniques that confirmed their successful preparation (Supplementary Figure S4B‐H, Supplementary Table S2). The amount of encapsulated doxorubicin in NP‐DOX‐PEI‐miR‐99a‐HA (70 mg/g SiO₂) was determined by thermogravimetric analyses and high‐performance liquid chromatography. Besides, we verified that doxorubicin was released in lysosomal extract, whereas there was a very low delivery in phosphate‐buffered saline (Supplementary Figure S4I), thus indicating that nanoparticles would deliver their content in lysosomes after being uptaken by cells. Moreover, the capacity of NP‐DOX‐PEI‐miR‐99a‐HA to specifically target CD44 was confirmed by a competition assay. Pretreatment of MDA‐MB‐231 cells with free HA inhibited the uptake of nanoparticles (60.5%, P < 0.001) (Figure 1K, Supplementary Figure S4J), thus evidencing the NP‐DOX‐PEI‐miR‐99a‐HA targeting to CD44.

We next examined the in vivo efficacy of the designed nanoparticles in a murine orthotopic xenograft model of triple‐negative BC (TNBC). NP‐DOX‐PEI‐miR‐99a‐HA induced the highest inhibition of tumor growth and cell proliferation; NP‐PEI‐HA and NP‐PEI‐mir‐99a‐HA displayed no significant differences with control, and NP‐DOX‐PEI‐HA showed a moderate capacity to inhibit tumor growth (Figure 1L‐M, Supplementary Figure S5A). Two days after last treatment, the tumor volume inhibition rates by NP‐DOX‐PEI‐HA and NP‐DOX‐PEI‐miR‐99a‐HA were 36.3% and 59.5%, respectively. Furthermore, the functional effect of miR‐99a‐5p was confirmed as COX‐2 and ABCG2 were downregulated in tumors treated with NP‐DOX‐PEI‐miR‐99a‐HA (Figure 1N). Biodistribution studies evidenced the ability of nanoparticles to target tumors, yet were also found in the kidneys, lungs, spleen, and liver (Supplementary Figure S5B). Moreover, nanoparticles had no toxicity in vivo (Supplementary Figure S5C‐D).

Cardiotoxicity is the major adverse effect of doxorubicin [10]. To assess the potential benefit of doxorubicin encapsulation in combination with miR‐99a‐5p, doses of free doxorubicin and NP‐DOX‐PEI‐miR‐99a‐HA with similar therapeutic efficacy were selected. Mice were treated with 1.6 mg/kg of free doxorubicin, whereas the equivalent dose of doxorubicin in nanoparticles was approximately 0.7 mg/kg, (44.0% of free drug) (Supplementary Figure S5E). Besides, higher levels of the marker of cardiac atrophy myh7 (myosin heavy chain 7) (Figure 1O) and the presence of cytoplasmic vacuoles in cardiac tissue (Supplementary Figure S5F) evidenced that free doxorubicin induced cardiotoxicity, while NP‐DOX‐PEI‐miR‐99a‐HA showed no differences with the control. These experiments proved the nanoparticles’ ability to diminish the required dose of doxorubicin and the consequent cardiotoxicity.

In conclusion, we have shown that miR‐99a‐5p modulated doxorubicin resistance in TNBC. To translate our findings, we designed a nanoparticle for specific co‐delivery of miR‐99a‐5p and doxorubicin that demonstrated its therapeutic potential in vivo, being more effective and less toxic than free doxorubicin alone.

CONFLICT OF INTERESTS

The authors declare no conflict of interests.

FUNDING INFORMATION

This work was supported by Spanish Government and co‐financed by FEDER Funds (PI18/01219, PI21/01351, PI18/00382, and RTI2018‐100910‐B‐C41), CIBER‐BBN (CB07/01/2012), CIBERONC (CB16/12/00481) and the Generalitat Valenciana (project PROMETEO 2018/024). I.G.‐C. was funded by Generalitat Valenciana (ACIF/2016/030), A.A. and A.L. were funded by Asociación Española Contra el Cancer. J.F.B., was funded by Instituto de Salud Carlos III and the European Social Fund for the financial support “Sara Borrell” (CD19/00038), V.C.‐N., was funded by Ministerio de Ciencia e Innovación (FPU grant), and J.M.C. was funded by Sociedad Española de Oncología Médica (Río Hortega‐SEOM).

AUTHOR CONTRIBUTIONS

IGC: Conceptualization, methodology, data analysis, writing—original draft preparation. AAA, AL (Ana Lameirinhas), JFB, VCN, FR, SZ, JMG: Methodology, data analysis. AL (Ana Lluch), BB: Conceptualization, funding acquisition. FS: data acquisition and analysis. JMC, RMM, PE: Supervision, conceptualization, writing—reviewing and editing, funding acquisition.

ETHICS APPROVAL AND CONSENT TO PARTICIPATE

All animal experiments were approved by the Institutional Review Board of Biomedical Research Institute INCLIVA (Valencia, Spain) (2020/VSC/PEA/0131).

CONSENT FOR PUBLICATION

Not applicable.

Supporting information

Supporting Information

DATA AVAILABILITY STATEMENT

The raw data required to reproduce these findings are available from the authors upon request.