Cancer impacts microRNA expression, release and function in cardiac and skeletal muscle

Abstract

Circulating microRNAs are emerging as important biomarkers of various diseases including cancer. Intriguingly, circulating levels of several microRNAs are lower in cancer patients compared with healthy individuals. In this study, we tested the hypothesis that a circulating microRNA might serve as a surrogate of the effects of cancer on microRNA expression or release in distant organs. Here we report that circulating levels of the muscle-enriched miR-486 is lower in breast cancer patients compared with healthy individuals, and that this difference is replicated faithfully in MMTV-PyMT and MMTV-Her2 transgenic mouse models of breast cancer. In tumor-bearing mice, levels of miR-486 were relatively reduced in muscle, where there was elevated expression of the miR-486 target genes PTEN and FOXO1A and dampened signaling through the PI3K/AKT pathway. Skeletal muscle expressed lower levels of the transcription factor MyoD which controls miR-486 expression. Conditioned media (CM) obtained from MMTV-PyMT and MMTV-Her2/Neu tumor cells cultured in vitro was sufficient to elicit reduced levels of miR-486 and increased PTEN and FOXO1A expression in C2C12 murine myoblasts. Cytokine analysis implicated TNFα and four additional cytokines as mediators of miR-486 expression in CM-treated cells. Since miR-486 is a potent modulator of PI3K/AKT signaling and the muscle-enriched transcription factor network in cardiac/skeletal muscle, our findings implicated TNFα-dependent miRNA circuitry in muscle differentiation and survival pathways in cancer.

Introduction

Extracellular/circulating microRNAs (miRNAs) have emerged as minimally invasive biomarkers of cancer progression and therapeutic response ^1–3. Imbalance in circulating miRNAs goes beyond cancer, as there is evidence for altered circulating miRNAs in Atherosclerosis and Alzheimer disease ^{4, 5}. Because of relative stability of these circulating miRNAs, the sera miRNA profiling has been suggested to be highly sensitive screening assay for early detection of various diseases ⁶.

The source of circulating microRNAs, particularly in cancer, remains an enigma as levels of several of circulating miRNAs show opposing pattern in tumor and in circulation ⁷. While tumor itself or circulating tumor cells are potential sources of miRNAs that are elevated in the sera/plasma of cancer patients, consistent observation of lower circulating levels of specific miRNAs in cancer patients compared with healthy controls suggest that systemic effects of cancer is causing overall changes in expression/release of miRNAs from distant organs ^8–10. For example, a recent study evaluating sera miRNA as a potential risk biomarker of breast cancer using prospectively collected sera from Sister Study Cohort showed down-regulation of five miRNAs in the sera of women who developed breast cancer ¹¹. Another report using breast tumors and sera from Asian Chinese patients showed down-regulation of miRNA in the sera of cancer patients ⁷. Our recent study provided a hint to the contribution of secondary organs in cancer-associated circulating miRNA changes as we observed elevated U6 small RNA in the sera of breast cancer patients who are clinically disease-free compared with healthy controls ¹². We proposed that cancer-induced epigenomic changes in distant organs cause elevated expression and release of U6 from these organs. However, this possibility has not been experimentally verified and the underlying mechanisms are unknown.

The goals of this study were to identify miRNAs that are present at a lower level in circulation in breast cancer models and then to elucidate mechanisms responsible for reduced levels of specific circulating miRNAs. We used two transgenic mammary tumor models; one is an aggressive tumor model and the other with relatively longer latency, to ensure that the results obtained are not unique to a specific model. Our results reveal specific deregulation in the expression of cardiac/skeletal muscle-enriched miRNA miR-486 in mammary tumor models. In vitro studies identified TNFα as a potential cancer-induced factor responsible for deregulation of miR-486 expression.

Methods

Human serum sample processing and miRNA extraction

The Indiana University Institutional Review board approved the use of human sera samples. Susan G. Komen for the Cure Normal Breast Tissue Bank at the IU Simon Cancer Center collected patient sera samples along with healthy volunteer controls after obtaining informed consent. All samples were collected in accordance with standard operating procedure described in the tissue bank website. MiRVana kit was used to isolate miRNA from 250 μL of sera (Applied Biosystems, Foster City, CA, USA). Sera were spiked with synthetic C. elegans miR-39 mimic (Qiagen, Valencia, CA, USA) before miRNA extraction and miR-486 expression was normalized to spiked miR-39 mimic levels. Characteristics of healthy controls and patients studied have been described in our previous publication ¹².

Transgenic Models of Breast Cancer

National Institutes of Health regulations regarding the use and care of experimental animals were followed while conducting animal studies and the study was approved by the Indiana University School of Medicine animal use committee. Male MMTV-PyMT or MMTV-Her2/Neu mice on a FVB/N background were randomly bred with normal FVB/N females to obtain female heterozygous for the PyMT and Her2/Neu oncogene. MMTV-PyMT and MMTV-Her2/Neu mice have been described previously ^{13, 14}. Neu oncogene used in this transgenic model is an activated form with 16 amino acids in-frame deletion of the extracellular domain ¹⁴. Blood, heart, and muscle were collected for miRNA preparation at the age of three months and five months from MMTV-PyMT and MMTV-Her2/Neu mice, respectively. As we have reported previously, MMTV-PyMT mice at this age have extensive tumor burden accompanied with metastasis to lungs ¹⁵. MMTV-Her2/Neu mice also develop lung metastasis by five months age, although their tumor burden and metastasis are not as extensive as in PyMT mice (¹⁴, data not shown). All animals had tumors at the time of tissue harvest. The age matched normal female mice were used as controls.

Quantitative reverse transcription-polymerase chain reaction (qRT-PCR)

Five μL of miRNAs (for sera) or 100 ngs (for tumor and normal mammary gland) were reverse transcribed into cDNAs in a final volume of 30 μL using a Taqman miRNA reverse transcription kit (Applied Biosystems, Grand Island, NY, USA). qPCR was performed using Taqman universal PCR mix (Applied Biosystems) and specific primers. Primers for U6 (#001973), miR-486 (#001278), miR-202 (#001195), and miR-30d (#000420) were purchased from Applied Biosystems. Each amplification reaction was performed in duplicate in a final volume of 20 μl with two μl of cDNA. qPCR reactions of sera from healthy subjects and patients with metastatic breast cancer for a particular probe were in the same plate to limit mechanical errors. The expression levels of miR-486 were normalized to miR-202 (mouse sera and mammary gland), U6 (cardiac and skeletal muscle), and miR30d (C2C12 cells) using 2- ^ΔΔCt method. In case of C2C12cells, one microgram of miRNA preparation was used for cDNA synthesis. mRNAs from C2C12 cells were prepared using RNAeasy kit from Qiagen and subjected to qRT-PCR using the primers that amplify both Ank1 and sAnk1 (primer sequences; forward: 5′ GAC GCA TGA CCT ACA GTC TTC 3′ and reverse: 5′ GCT ATC CTC TCC CTT CTT CTC T 3′) and βActin (forward: 5′ AAT GAG GCC GAG GAC TTT GAT TGC 3′ and reverse: 5′ AGG ATG GCA AGG GAC TTC CTG TAA 3′).

MicroRNA array

MicroRNA profiling was performed using the Taqman MicroRNA array A that included an assay plate containing 384 probes of rodent miRNAs (#4398979, Applied Biosystems). PCR was conducted as per instructions from the manufacturer using nine μl of pre-amplified cDNAs. Pre-amplification reactions were done as per instructions from Applied Biosystems using the TaqMan MicroRNA RT Kit (part no. 4366596). The sera from four each of control, MMTV-PyMT, and MMTV-Her2/Neu were used for the array. Probes that showed undetectable signals were given a CT value of 40 for calculation. Supplementary Table 1 (Table S1) provides CT values for each of the probes in all 12 samples. Normalization using miR-202 was done using the 2- ^ΔΔCt method.

Cell culture and conditioned media

Mouse myoblast C2C12 cells were seeded in 6-well plates (5x10⁵ cells per well) in DMEM plus 10% FBS overnight. Mammary tumor cells generated from MMTV-Her2/Neu ¹⁶ and MMTV-PyMT mice ¹⁵ were cultured overnight in the same media and changed to serum-free DMEM medium for 24 hours. C2C12 cells were treated with CM for 24 hours. For neutralizing antibody assay, CM was pre-incubated with one μg/ml of anti-TNFα antibody (R&D systems, Minneapolis, MN) at room temperature for one hour before adding to C2C12 cells. To directly measure the effects of cytokines on miR-486 expression, C2C12 cells were treated with 20 ng/ml of CCL2, IFNγ, IL-1α, or TNFα (R&D systems) overnight. For promoting differentiation to myotubes, 5000 C2C12 cells were plated in 6-well plates and maintained in 2% of horse serum containing media for seven days. Serum-free control or MMTV-PyMT or MMTV-Her2/Neu tumor cell line-derived CM were added two days after plating.

Western blotting

After indicated treatments, cells were washed in PBS and lysed in RIPA buffer with protease/phosphatase inhibitors (Sigma, St. Louis, MO, USA). Thirty micrograms of proteins were used for western blotting. Antibodies against FOXO1A (Upstate, Lake Placid, NY, USA), βActin (Sigma), pAKT (Cell Signaling Technologies, Danvers, MA), p27 (BD Biosciences, San Jose, CA), PTEN, and MyoD (Santa Cruz Biotechnologies, Santa Cruz, CA, USA) were used for western blot analyses as per instructions from manufacturers.

Cytokine array and ELISA

Serum-free CM derived from MMTV-Her2/Neu and MMTV-PyMT cultures were subjected to immunoblotting based cytokine profiling using the mouse cytokine array panel A (#ARY006, R&D Systems). ELISA was used to measure TNFα levels in mouse sera and CM (BioLegand, San Diego, CA for mouse).

Statistical analysis

Expression levels of sera miRNAs were compared using ANOVA. A p value <0.02 was considered statistically significant in array experiments and <0.05 in validation experiments. miR-486 in human sera was evaluated using ANOVA.

Results

Lower levels of circulating miR-486 in transgenic mice with mammary tumors and in breast cancer patients

To better understand the mechanism of lower circulating levels of specific miRNAs in cancer patients compared with healthy controls, we profiled miRNAs in the sera of controls and transgenic mice with mammary tumors derived upon Her2/Neu (MMTV-Her2/Neu) or polyoma middle-T-antigen (MMTV-PyMT) oncogene overexpression ^{13, 14}. Because of disagreements regarding normalization controls for such studies and each study claiming different miRNAs as better normalization controls ^{7, 17}, data were analyzed without normalization first. Additionally, analysis was conducted by selecting a miRNA, which showed highest stability value across samples, as a normalization control similar to the previously published Sister Study cohort report ¹¹. Principle Component Analysis (PCA) showed a near perfect separation of three groups when data were analyzed without normalization compared with normalization using miR-202 (Fig. 1A and B). Reason for the discrepancy in separation of samples to three groups with and without normalization is unclear. However, a recent study has shown PCA as a better method to identify circulating microRNAs ¹⁸. Samples utilized in this study were from inbred mice housed under similar condition, and collected and analyzed at the same time, which should limit inter-mice and technical variability. Therefore, extensive additional studies are required to find ideal normalization control. Alternatively, when another miRNA or small RNA is used as a normalization control, it may be ideal to present results as a ratio between test miRNA and specific normalization control RNA as we have presented below. Detailed miRNA profiles with CT values and comparison between three groups with and without normalization are shown in supplementary Table 1 (Table S1).

Figure 1. The sera of transgenic mice with mammary tumors display distinct miRNA profile.

A) Principle Component Analysis (PCA) of miRNAs in sera of control, MMTV-Her2/Neu, and MMTV-PyMT mice without normalization. B) PCA of miRNAs in sera of control, MMTV-Her2/Neu, and MMTV-PyMT mice after normalization with miR-202. C) qRT-PCR analysis confirmed down-regulation of miR-486 in sera of transgenic animals with mammary tumors. D) Analysis of circulating miR-486 levels in healthy and metastatic patients using miRNA preparations after spiking sera with C. elegans miR-39 to correct for technical variability. E) miR-486 levels in the normal mammary gland (n, 7–8), MMTV-Her2/Neu (n=6) and MMTV-PyMT-derived mammary tumors (n=10).

With p value cut off of <0.02 and no normalization, we observed 16 miRNAs being present differentially in the sera of MMTV-Her2/Neu mice compared with control mice with two of them being down-regulated (Table 1). In contrast, the sera of MMTV-PyMT mice contained elevated levels of 34 miRNAs (Table 1). Difference in number of circulating miRNAs between MMTV-PyMT and MMTV-Her2/Neu is statistically significant (P=0.01, Fisher’s exact test, two-tailed). Despite shorter latency, tumors and lung metastasis were more advanced in MMTV-PyMT transgenic mice compared with MMTV-Her2/Neu mice, which may be a reason for significantly higher number of circulating miRNA changes in MMTV-PyMT mice compared with MMTV-Her/Neu mice. Expectedly, unique changes were more common in the MMTV-PyMT mice (26 out of 34) compared with the MMTV-Her2/Neu mice (eight out of 16).

Table 1. Significantly altered circulating miRNAs in transgenic mice with mammary tumors.

No normalization was used in this analysis.

	MMTV-Her2/Neu vs. control		MMTV-PyMT vs. control
miRNA	P value	Fold change	P value	Fold change
MMTV-Her2/Neu specific
miR-486	0.0001	−6.54	0.72	−1.11
miR-7a	0.0025	58.47	0.69	−1.48
miR-743a	0.007	2.03	1	−1
miR-381	0.0079	2.66	1	−1
miR-139-3p	0.015	88.62	0.03	46.2
miR-129-3p	0.016	35.61	0.492	2.39
miR-191	0.018	−4.45	0.21	2.00
miR-34a	0.019	40.757	0.606	2.00
MMTV-PyMT specific
miR-136	0.978	1.01	2.19E-06	81.41
miR-202-3p	0.537	−1.15	0.00046	3.35
miR-574-3p	0.50	1.34	0.00094	7.92
miR-146b	0.271	2.15	0.001	21.47
miR-134	0.0248	6.4	0.0015	21.7
miR-667	0.668	1.18	0.0017	5.28
miR-685	0.633	1.43	0.002	21.7
miR-24	0.03	4.2	0.003	8.9
miR-132	0.019	49.99	0.0038	209.42
miR-542-5p	0.45	2.6	0.005	97.17
miR-324-3p	0.11	4.05	0.006	17.07
miR-223	0.919	−1.06	0.0066	8.17
miR-92a	0.718	−1.15	0.007	3.94
miR-342-3p	0.08	2.73	0.007	5.83
miR-410	0.857	1.31	0.0096	127.559
miR-200a	0.038	7.07	0.01	13.62
miR-484	0.639	−1.23	0.01	4.05
miR-139-5p	0.715	1.255	0.01	6.96
miR-429	0.035	53.58	0.01	173.95
miR-187	0.94	1.07	0.01	19.58
miR-30a	0.141	2.67	0.014	6.37
miR-511	0.031	15.93	0.015	25.74
miR-423-5p	0.51	−1.52	0.016	6.09
miR-335-5p	0.364	5.14	0.016	155.23
miR-146a	0.303	1.93	0.016	5.9
miR-125a-3p	0.238	13.7	0.019	359.00
Commonly deregulated in both transgenic mice
miR-193b	0.009	6.7	0.0001	36.2
miR151-3p	0.0072	5.199	0.0003	13.73
miR-27b	0.007	171.86	0.0004	3013.24
miR-671-3p	0.0095	47.41	0.0013	220.48
miR-200b	0.014	9.82	0.018	8.7
miR-183	0.016	23.44	0.004	56.17
miR-210	0.02	6.53	0.00082	27.37
miR-132	0.02	49.99	0.0038	209.4

When miRNA profiles were normalized using miR-202, only four miRNAs were present differentially in the MMTV-Her2/Neu sera compared with the sera from non-transgenic mice with one of them being lower, whereas five miRNAs were present differentially in the sera of MMTV-PyMT with four of them being down-regulated (Table 2). However, number of circulating miRNAs differentially present in the MMTV-PyMT may be an underestimation in this analysis because of overall increase in miR-202 levels in MMTV-PyMT sera but not in MMTV-Her2/Neu sera compared with controls (Table 1). miR-146b levels were elevated, whereas miR-486 levels were lower in one or both transgenic mice models giving us the confidence that these two miRNAs are the major differentially expressed/secreted miRNAs in cancer.

Table 2. Significantly altered circulating miRNAs in transgenic mice with mammary tumors.

Data were analyzed using miR-202 as a normalization control. Although miR-202 was present in all sera samples analyzed and showed least variability between samples, its levels were higher in the sera of MMTV-PyMT mice compared with control or MMTV-Her2/Neu mice.

	MMTV-Her2/Neu vs. control		MMTV-PyMT vs. control
miRNA	P value	Fold change	P value	Fold change
MMTV-Her2/Neu specific
miR-204	0.009	8.9	0.507	2.66
miR-375	0.017	2.82	0.712	−1.31
miR-381	0.02	3.2	0.375	−3.63
MMTV-PyMT specific
miR-450b-5p	0.43	1.18	0.01	−3.63
miR-142-3p	0.119	−2.02	0.01	−12.30
miR-150	0.073	−1.89	0.014	−3.41
miR-146b	0.397	2.59	0.018	6.16
Common to both
miR-486	0.0006	−5.69	0.001	−3.69

We selected miR-486 for further study because its expression is enriched in cardiac and skeletal muscle, which represents ~40% of body mass ¹⁹. In addition, CT values ranged from 15–20 suggesting that its levels can be reliably measured. We verified the microarray results by qRT-PCR to confirm cancer-specific down-regulation of circulating miR-486 in both MMTV-Her/Neu and MMTV-PyMT mice using miR-202 as a normalization control (Figure 1C). As with microarray, in these validation experiments, miR-486 levels were lower in sera of MMTV-Her2/Neu mice (3.78±0.44 in control versus 1.55±0.44 in MMTV-Her2/Neu, p=0.0004) but not in sera of MMTV-PyMT mice when data were analyzed without normalization, further confirming reproducibility (data not shown).

To confirm the relevance of data obtained in animal models to human, we measured miR-486-5P levels (equivalent of mmu-miR-486) levels in the sera of breast cancer patients with metastasis (N=17) and healthy women (N=16). Circulating miR-486 levels were lower in patients with breast cancer metastasis compared with healthy (fold change −3.14, p=0.002 ANOVA, metastasis versus normal) (Figure 1D).

To investigate whether there is any relationship between miR-486 in tumor and in circulation, we measured miR-486 levels in normal mammary gland and tumors from both transgenic mice. While PyMT tumors and normal mammary gland expressed similar levels of miR-486, Her2 tumors showed elevated miR-486 expression compared with normal (Figure 1E). Therefore, differential levels of circulating miR-486 in three groups are less likely due to altered expression in tumors compared with normal mammary gland. We do acknowledge limitations of this analysis because normal mammary gland and tumors differ in their epithelial content. Similar analysis of public databases for miR-486 in breast cancer and normal breast gave ambiguous results. While TGCA dataset ²⁰ showed lower mir-486 in breast tumors compared with normal breast, three other datasets (GSE32922, GSE44124 and GSE53179;^{21, 22}) failed to demonstrate reduced miR-486 in tumors (supplementary Figure S1A). Moreover, in the TCGA dataset, higher miR-486 expression was associated with worst outcome, which is not compatible with the observation of lower tumor-specific expression (Figure S1B). Although reduced expression of miR-486 in breast tumors leading to lower circulating miR-486 cannot be completely ruled out, based on the results in murine models and ambiguity of results obtained with breast tumors and normal breast tissue, we favor the alternative possibility of systemic effects of cancer causing lower circulating miR-486 levels.

Cardiac and skeletal muscle of transgenic mice with mammary tumors express lower miR-486 compared with controls

To determine whether lower circulating miR-486 in transgenic mice with cancer can be attributed to its lower expression in muscle and heart, we measured miR-486 in miRNA preparation from these organs. Animals used in these experiments are different from those used for sera miRNA analysis. Indeed, miR-486 levels were lower in the heart and muscle of MMTV-Her2/Neu transgenic mice (n=6) and in the heart of MMTV-PyMT mice compared with control mice (n=6 for control, n=5 for muscle, n=4 for heart, Figure 2A). U6 is an appropriate control for heart and muscle tissue based on a previous study and our finding that it was not significantly different between PyMT, Her2, and control mice ¹⁹.

Figure 2. Cancer cell-derived soluble factors reduce miR-486 expression in muscle and increase miR-486 target proteins.

A) Heart of both MMTV-Her2/Neu and MMTV-PyMT and skeletal muscle of MMTV-Her2/Neu mice with mammary tumors expressed lower levels of miR-486 compared with healthy controls (N=6). B) CM from MMTV-Her2/Neu and MMTV-PyMT tumor-derived cell lines reduced miR-486 expression in C2C12 cells. CM pre-heated at 80 C partially lost the ability to reduce miR-486 expression. Asterisks and dollar sign indicate statistically significant differences. C) CM from MMTV-Her2/Neu and MMTV-PyMT tumor-derived cell lines reduced Ank1 mRNA. D) CM from MMTV-Her2/Neu and MMTV-PyMT tumor-derived cell lines up-regulated miR-486 target proteins in C2C12 cells.

Conditioned media (CM) from MMTV-Her2/Neu and MMTV-PyMT mammary tumor cell lines reduce miR-486 in C2C12 cells

To delineate the mechanism involved in cancer-induced changes in miR-486 expression in muscle, we utilized undifferentiated murine myoblast C2C12 cell line as a model system. The effects of CM from tumor cell lines derived from MMTV-Her2/Neu and MMTV-PyMT mice on miR-486 expression in these cells were measured. The CM from both MMTV-Her2/Neu and MMTV-PyMT tumor-derived cell lines reduced miR-486 levels in this myoblast cell line (Fig. 2B). The ability to inhibit miR-486 expression was reduced partially when CM was pre-treated for 20 minutes at 80 C indicating that a protein factor(s) in the CM is repressing miR-486 expression. miR-30d expression was used as a normalization control. However, results were similar when data were analyzed without normalization in majority of experiments (data not shown).

mir-486 is transcribed from the intron 40 of Ankyrin-1 (Ank1) gene, which encodes for an ankyrin repeat protein and an erythroid specific enhancer-promoter controls its expression in erythroid cells ¹⁹. However, muscle cells express a smaller Ank1 (sAnk1) transcript containing exon 39a and exons 40–42 utilizing an alternative promoter immediately upstream of exon 39a. sAnk1 and miR-486 are expressed coordinately in muscle cells and muscle-enriched transcription factors MyoD and myocardin-related transcription factor-A (MRTF-A) control their expression ¹⁹. Since CM from tumor-derived cell lines reduced miR-486 in C2C12 cells, we examined the effects of CM on Ank1 expression. Similar to miR-486, the expression of Ank1 was reduced in CM treated cells and pre-heat treated CM was less efficient in reducing Ank1 expression (Figure 2C). Please note that RNAs for these experiments were prepared from different batch of cells using independent methods and βactin was used as a control for normalization in Ank1 expression analysis. Thus, two independent assays demonstrate an effect of cancer cell-derived factors on the expression of miR-486.

FOXO1A and PTEN are the well-established targets of miR-486. Therefore, by down-regulating PTEN, miR-486 activates PI3K/AKT survival pathway in cardiac and skeletal muscle ¹⁹. In addition, miR-486 has been shown to suppress muscle wasting by targeting FOXO1 ²³. Since tumor cell line-derived CM reduced miR-486 levels, we determined the expression levels of its target proteins in C2C12 cells with and without CM treatment. CM pre-treated at 80 C for 20 minutes was used a control. CM from both MMTV-Her2/Neu and MMTV-PyMT cell lines increased the levels of FOXO1A and PTEN with concomitant decline in pAKT levels (Figure 2D). The levels of FOXO1A target protein p27 were elevated in cells treated with CM from MMTV-Her2/Neu and MMTV-PyMT tumor cell lines ²⁴. None of these changes were observed when cells were incubated with heat-treated MMTV-PyMT cell line-derived CM. For unknown reason, heat-treated MMTV-Her2/Neu tumor-derived cell line CM was still able to increase PTEN and p27 but not FOXO1A. Nonetheless, these results provide evidence for a factor(s) derived from cancer cells in reducing miR-486 levels and as a consequence, increasing miR-486 targets in the myogenic cell line.

Tumor Necrosis Factor Alpha (TNFα), secreted by cancer cells, alters miR-486 expression in C2C12 cells

We used cytokine arrays to identify cytokines present in the CM from both cell lines. This array measures 40 different cytokines. Although each CM contained unique cytokines (TIMP1 in case of PyMT and G-CSF, CCL-1, CCL-5, CXCL1, CXCL2, CXCL10, and IL-1ra in case of Her2/Neu), GM-CSF, TNFα, CCL-2, IFNγ, and IL-1α were the common cytokines secreted by both tumor lines (Figure 3A). The ability of some of these cytokines to alter miR-486 and Ank1 expression was examined. Although all four cytokines tested reduced miR-486 expression, only IL-1α and TNFα reduced both miR-486 and Ank1 expression (Figure 3B and C). These results suggest that while TNFα and IL-1α reduce miR-486 by targeting Ank1 regulatory regions, CCL-2 and IFNγ regulate miR-486 expression and/or maturation independent of Ank1 gene transcription.

Figure 3. TNFα mediates miR-486 repression.

A) Cytokine array identified GM-CSF, CCL2, IFNγ, IL-1α, and TNFα as common cytokines secreted by both MMTV-Her2/Neu and MMTV-PyMT tumor-derived cell lines. B) The effects of CCL2, IFNγ, IL-1α, and TNFα on miR-486 expression in C2C12 cells. C) The effects of CCL2, IFNγ, IL-1α, and TNFα on Ank1 expression in C2C12 cells. D) Neutralizing antibody against TNFα partially reversed the effects of CM from MMTV-Her2/Neu and MMTV-PyMT tumor-derived cell lines on miR-486 expression in C2C12 cells. E) Neutralizing antibody against TNFα reduced the effects of CM from MMTV-Her2/Neu and MMTV-PyMT tumor-derived cell lines on Ank1 expression in C2C12 cells. F) CM from MMTV-Her2/Neu and MMTV-PyMT tumor-derived cell lines contained TNFα. G) Sera of MMTV-Her2/Neu and MMTV-PyMT mice contained higher levels of TNFα compared with control mice.

We focused on TNFα because of its previously described role in reducing MyoD expression in muscle, an important transcription factor required for miR-486 expression, and in muscle dysfunction ^{19, 25}. Towards this end, we pre-treated CM with either control IgG or neutralizing antibody against TNFα and then applied to C2C12 cells. TNFα neutralizing antibody significantly prevented CM-mediated suppression of miR-486 and Ank1 expression (Figure 3D and E). Using ELISA assay, we confirmed the presence of TNFα in CM from MMTV-Her2/Neu and MMTV-PyMT cell lines and in sera of transgenic mice with tumors (Figure 3F and G). Similar analysis of sera from patients with metastatic breast cancer and healthy donors showed a trend of elevated circulating TNFα in cancer patients compared with healthy (Figure S1C).

CM from tumor cell lines reduce miR-486 expression in myotubes

The above studies were performed in undifferentiated C2C12 cells. In order to determine whether CM from tumor cell lines reduce miR-486 in myoblasts undergoing differentiation to myotubes, we grew C2C12 cells in media containing 2% horse serum ²⁶. CM from both cell lines reduced the levels of miR-486 in the differentiated cells (Fig. 4). We also noted lower cell density when cells were treated with tumor cell-derived CM compared with control CM, possibly indicating an effect of CM on miR-486-mediated cell survival pathway (Figure 4).

Figure 4. CM from tumor cell lines decrease miR-486 expression in differentiated C2C12 cells.

C2C12 cells were allowed to undergo differentiation by replacing FBS with horse serum. Phase contrast images of myotubes under control or CM-treated conditions are shown. The levels of miR-486 with or without CM treatment are also shown.

Deregulation of PI3K-PTEN-AKT survival network in cardiac and skeletal muscle of transgenic mice

Although in vitro studies indicated a clear effect of factors secreted by cancer cells on miR-486 expression and survival signaling network in myoblasts, we wanted to confirm similar scenario in intact animals. Towards this end, we measured the levels of miR-486 target proteins in extracts from heart and muscle. Heart from transgenic animals contained lower levels of pAKT compared with control animals despite insignificant difference in PTEN levels between groups (Figure 5A). PTEN is highly abundant protein in heart, which may be a reason for not detecting differences in its levels between groups. In contrast, muscle of transgenic mice contained elevated PTEN and lower pAKT (Figure 5B). Since PI3K/pAKT has a cardio-protective function and prevents muscle atrophy by inhibiting FOXO transcription factors ²⁷, whereas elevated levels of phosphorylated/activated p38 kinase is associated with apoptosis and is often increased during ischemic heart disease ²⁸, we next measured the levels of phospho-p38 (T180/Y182). Indeed, extracts of heart from transgenic mice contained elevated phospho-p38 compared with extracts from control mice (Figure 5A). These results suggest that tumor-induced factors such as TNFα reduce the expression of miR-486 in heart leading to impairment in PI3K/AKT-dependent survival pathway and elevated p38 kinase pathway.

Figure 5. Heart and muscle of transgenic mice display defective PI3K/AKT pathway compared with control mice.

A) pAKT, PTEN, MyoD, and pp38 levels in the heart of mice. Quantitative differences are shown at the bottom. Extracts from four mice in each group were used. B) pAKT, PTEN, MyoD, and pp38 levels in the muscle of mice. C) Model depicting the effects of cancer on signaling events in heart and muscle leading to lower miR-486 in circulation.

Since sera of transgenic animals contained elevated levels of TNFα, which can impact miR-486 expression by targeting MyoD ¹⁹, we next examined MyoD protein levels in heart and muscle of control and transgenic animals. MyoD levels were lower in skeletal muscle of tumor bearing animals (Figure 5B). Similar trend was observed in heart, although differences did not reach statistical significance. In summation, our results suggest the effect of cancer on microRNA expression in distant organs with an impact on the survival-signaling network in these organs.

Discussion

A number of recent reports have described circulating miRNAs as potential biomarkers of cancer ^{7, 11, 17}. However, this field still suffers from lack of reproducibility, as there is a minimum overlap in cancer-specific circulating miRNAs identified in different studies. There are two possible explanations for this lack of reproducibility; first is the difficulty in finding a suitable normalization control and second is the recent realization of large scale transcriptome variation among healthy humans, which makes it difficult to assign “normal” value to circulating miRNAs ²⁹. Nonetheless, at least few specific miRNAs have been detected in disease conditions in more than one study. For example, elevated circulating miR-181a and miR-222 have been observed in breast cancer in more than one study ^{7, 11, 17}. miR-151-3p, miR-134, and miR-671-3p identified in our animal models have been shown to be elevated in the sera of cancer patients ^{7, 11}. Therefore, studies that combine analysis in transgenic animal models in a similar genetic background and in human samples, as done in this study, may identify a set of circulating miRNAs that can be used as biomarkers in a clinical setting.

Several miRNAs that are present at higher levels in the sera of transgenic mice have previously been shown to have oncogenic role. For example, miR-27b, which is elevated in the sera of both transgenic mice (Table 1), is a context-specific oncogene and both Her2 and TNFα increase its expression in breast cancer cells ³⁰. In fact, miR-27b was not measurable in the sera of control animals but readily detectable in the sera of transgenic mice (Table S1). Furthermore, combination of two circulating miRNAs, miR-27b and miR-15b, has been shown to discriminate non-small cell lung cancer patients from healthy controls ³¹. miR-210 is a hypoxia inducible oncogene, which predicts poor outcome in breast cancer patients ³². Consistent with rapid tumor progression, which often leads to hypoxia, MMTV-PyMT tumor bearing mice had much higher circulating miR-210 than MMTV-Her2/Neu tumor bearing mice (Table 1). Circulating levels of miR-210 and miR-200b correlate with the presence of circulating tumor cells in breast cancer patients ¹⁰. miR-146a and miR-146b elevated in the sera of MMTV-PyMT mice have previously been shown to target BRCA1 and are expressed at a higher level in basal-like and triple negative breast cancers ³³. In addition, circulating miR-146 may suggest overall inflammatory status as both miR-146a and miR-146b are NF-κB-inducible miRNAs ^{34, 35}. miR-183, which is elevated in the sera of both transgenic mice, is overexpressed in ductal carcinoma in situ compared with normal breast ³⁶. Thus, deregulation of certain miRNAs in breast/mammary tumors is common across species and independent of specific oncogenic events.

There have been limited attempts to understand why circulating levels of certain miRNAs are lower in cancer despite consistent observation of such a phenomena. We selected miR-486 as a model miRNA to study this aspect because it is a unique miRNA with no family members and is expressed predominantly in heart and muscle, which represent 40% of body mass ¹⁹. Unlike most other miRNAs with family members sharing targets and thus compensating for the loss of expression of a family member, loss of miR-486 expression is likely to have consequences. Consistent with this possibility, disease phenotypes are associated with reduced miR-486 expression as evident in the muscle of Duchene Muscular Dystrophy patients ³⁷. Physical exercise can have a negative influence on its release into circulation suggesting a link between muscle biology and circulating levels of miR-486 ³⁸. Our results have demonstrated an effect of mammary tumors on its expression in skeletal and cardiac muscle. Similar scenario may exist in other cancers as lower circulating miR-486-5p in relation to miR-21 or miR-126 is observed in lung cancer patients with poor outcome ³⁹.

To date, most research on cardiac health of cancer patients is focused on cardiac toxicity of cancer therapy ⁴⁰. Our study raises the possibility that cardiac dysfunction occurs during cancer progression. As presented schematically in Figure 5C, such a cancer-induced collateral damage to heart/muscle involves a miRNA network. Similarly, cancer-induced skeletal muscle dysfunction and cachexia is observed in at least 50% of patients although cachexia is rare in breast cancer patients ⁴¹. Recently, cachexia has been defined as a syndrome that progresses through various stages; precachexia to cachexia to refractory cachexia ⁴². Since tumor bearing transgenic mice did not show severe cachexia at the time of sacrifice but displayed some of the molecular defects associated with skeletal muscle dysfunction including lower MyoD expression, drop in circulating miR-486 levels may provide an indication to precachectic stage.

Neutralizing antibody against TNFα is already in clinical use for other diseases ⁴³. At least in animal models, TNFα neutralizing antibody reduced mammary tumor growth ⁴⁴. It may be worth considering these treatments not only to inhibit tumor growth but also to reduce side effects of cancer. The efficacy of a treatment in patients with metastasis is often measured by the ability of the drug to shrink metastasis. However, few of these drugs including anti-TNFα antibody may not be effective in reducing metastasis but effective in reducing cancer-induced collateral damage. Such treatments may help to extend and/or improve quality of life. Since RNA-based therapies are increasingly being developed as treatment with improved delivery system ^{45, 46}, replenishing miRNAs that are lower in circulation of cancer patients may be an option to restore cardiac and muscle function.