microRNA-148a Is a Prognostic oncomiR That Targets MIG6 and BIM to Regulate EGFR and Apoptosis in Glioblastoma

Glioblastoma is an extremely aggressive tumor that accounts for the majority of deaths due to primary brain neoplasms (1). Despite the most advanced treatment with combinations of surgery, radiotherapy, and chemotherapy, glioblastoma is associated with a median survival of only 14 months (2). Factors responsible for glioblastoma malignancy and poor prognosis include rapid cell growth, resistance against apoptosis, and distant invasion of the surrounding brain (1, 3).

Receptor tyrosine kinase (RTK) pathways are deregulated in the vast majority of glioblastomas (4, 5). Among RTKs, the EGF receptor (EGFR) is the most commonly altered (6). It is mutated and/or amplified in 40% and overexpressed in >60% of tumors (7, 8). Activation of EGFR induces tumor cell growth, migration, and invasion, as well as resistance to chemotherapy and radiation (6, 9). EGFR signaling and protein half-life are tightly regulated (10). Mitogen-inducible gene 6 (MIG6) regulates EGFR signaling and turnover by binding EGFR and directly inhibiting tyrosine kinase activity, increasing clathrin-dependent EGFR endocytosis and trafficking into the lysosome, and promoting EGFR degradation (11–13). Ablation of MIG6 induces tumor formation, supporting a tumor suppressor function of MIG6 (11, 14). The MIG6 gene is located on chromosome 1p36, which is subject to focal deletions in glioblastoma. The Cancer Genome Atlas (TCGA) data analysis showed that 15 out of 430 glioblastoma samples contain homozygous deletions in 1p36 (14) but that MIG6 expression is downregulated in approximately 50% of primary tumor samples and glioblastoma cell lines (11). Therefore MIG6 deletions only account for a small fraction of the glioblastoma tumors with reduced MIG6 expression.

Resistance to apoptosis is a big obstacle in glioblastoma therapy (15, 16). Apoptosis in the intrinsic pathway is regulated by the balance between proapoptotic (Bax, Bak, BIM, and Bad) and antiapoptotic (Bcl-2 and Bcl-xL) members of the Bcl-2 family (17). Proapoptotic BIM (BCL2L11) is localized to the mitochondria where it initiates the mitochondrial cell death pathway by directly activating Bax/Bak-dependent apoptosis. BIM has been shown to be an important mediator of targeted therapy-induced apoptosis in solid tumors. BIM is downregulated in 29% of glioblastoma cases based on TCGA analysis (18, 19). However, the causes of BIM downregulation in glioblastoma are not known.

microRNAs (miRNA) are short noncoding RNA molecules that regulate gene expression by binding to the 3′ untranslated region (3′UTR) of target mRNA and inducing mRNA degradation and/or inhibition of protein synthesis (20, 21). Deregulation of miRNA expression has been associated with cancer formation through alterations in either oncogenic or tumor suppressor gene targets (20, 22). A number of miRNAs are deregulated in glioblastoma and play important roles in tumor formation and growth (23–31). However, a role for miR-148a in glioblastoma has not been described before.

We analyzed miRNA expression in >500 patient glioblastomas in the TCGA database and found that miR-148a is upregulated and predicted poor patient survival. We therefore embarked on a comprehensive study of miR-148a in glioblastoma. Our data show for the first time that miR-148a is upregulated in glioblastoma, where it exerts oncogenic effects in vitro and in vivo by regulating BIM, MIG6, and EGFR. MiR-148a is therefore a novel oncomiR and potential therapeutic target in glioblastoma.

Glioblastoma cell lines U87, U373, A172, T98G, SNB-19, and U251 were from American Type Culture Collection, who authenticates cell lines with short tandem repeat (STR) profiling. Cells lines that were used for more than 6 months after purchase were reauthenticated by STR profiling in 2013 by Laragen, Inc. Glioblastoma stem cells (GSC) 1228, 0802, and 0308 (a kind gift from Dr. Jeongwu Lee, Cleveland Clinic, Cleveland, OH) were isolated from patient surgical specimens and characterized for tumorigenesis, pluripotency, self-renewal, stem cell markers, and neurosphere formation (32). Glioblastoma surgical specimens (n = 18) and normal brain (n = 7) were obtained from patients undergoing surgery at the University of Virginia Hospital (Charlottesville, VA) according to protocols approved by the Internal Review Board.

The collection of data from TCGA was compliant with all laws and regulations for the protection of human subjects, and necessary ethical approvals were obtained. Analysis of all data was done in the R project (33). For analysis of differential expression and determination of the effects of miRNA (miR)-148a on patient survival, Agilent 8 × 15 k miRNA expression for 491 glioblastoma and 10 normal unmatched brain samples was downloaded along with clinical information from the TCGA database [level 2 (normalized) data, November 2012]. Cox regression analysis of all samples with miRNA and survival data (n = 482) was performed to determine whether miR-148a levels were a risk indicator for survival. The expression of miR-148a was also compared in normal brain (n = 10) with glioblastoma (n = 491) using the R-based Limma package (34).

miScript Primer Assay Hs-miR-148a was used for measuring miR-148a. Total RNA was extracted from glioblastoma cell lines and GSCs. RNA samples were reverse transcribed using the miScript Reverse Transcriptase Kit (QIAGEN), and quantitative real-time PCR (qRT-PCR) analysis was performed using the 7500 Real-time PCR System (Applied Biosystems). qRT-PCR was also used to assess the mRNA levels of MIG6 and BIM. The primer sequences were: MIG6-forward: 5′-GACAATTTGAGCAACTTGACTTGG-3′, MIG6-reverse: 5′-GGTTACTTAGTTGTTGCAGGTAAG-3; BIM-forward: 5′-TGGCAAAGCAACCTTCTGATG-3′ and BIM-reverse: 5′-GCAGGCTGCAATTGTCTACCT-3′. Human U6B and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) primers (QIAGEN) were used as controls.

Glioblastoma cells and GSCs were transfected with 20 nmol/L pre–miR-148a, anti–miRNA-148a, or control-miR (Ambion), using Oligofectamine or Lipofectamine RNAimax (Invitrogen) according to the manufacturer's instructions. Plasmid transfections were performed with Fugene 6 (Roche). miR-148a expression was verified by qRT-PCR 72 hours and 7 days posttransfection.

Lentiviruses encoding the pEZX-AM04 expression cassette containing a hygromycin resistance gene as well as the antisense sequence for miR-148a and the red fluorescent protein mCherry gene under the U6 promoter (pEZX-AM04; GeneCopoeia; Supplementary Fig. S2B and S2C) were generated with pPACKH1 Lentivector packaging Plasmid mix (System Biosciences) and concentrated using PEG-it Virus Precipitation Solution (System Biosciences.). U87 cells were infected with the lentiviruses or control viruses lacking the anti–miR-148a sequence. After culturing in selection media, mCherry was detected by fluorescence microscopy. A stable infection efficiency of approximately 100% was attained (Supplementary Fig. S2A).

For growth, glioblastoma cells and GSCs were transfected with pre–miR-148a, anti–miR-148a, or control. Three days posttransfection, the cells were counted for 5 days with a hemocytometer. For apoptosis, cells were transfected as above and Annexin V-PE/7AAD flow cytometry was used to determine the dead and apoptotic cell fractions as previously described (35).

The effects of miR-148a expression on cell migration and invasion were assessed using the wound-healing and Transwell assays as previously described (36).

GSCs were grown in low EGF and fibroblast growth factor medium (20 ng/mL each) and transfected with either anti- or pre–miR-148a or controls for 72 hours. The cells were dissociated into single cells in buffer (EDTA 1 mmol/L, bovine serum albumin 0.5% in PBS) and 1,000 single cells were incubated for 7 days. Secondary neurospheres containing more than 30 cells were counted.

Tumor xenografts were generated by implantation of 1228 GSCs transfected with anti–miR-148a and U87 cells engineered to stably express anti–miR-148a. 1228 (1 × 10⁵ cells; n = 6) and U87 cells (3 × 10⁵ cells; n = 10) were stereotactically implanted into the striata of immunodeficient mice. Four weeks after tumor implantation, the animals were subjected to brain MRI. To measure the tumor size, 30 μL of gadopentetate dimeglumine (Magnevist, Bayer Healthcare) was intraperitoneally injected 15 minutes before scanning and tumor volume was quantified as previously described (37, 38).

Immunoblotting was performed as previously described using antibodies for MIG6 (Santa Cruz Biotechnologies), BIM, EGFR, and p-EGFR (Cell Signaling Technology). All blots were stripped and reprobed with β-actin or GAPDH (Santa Cruz Biotechnologies) as control. Blots in which differences were not obvious were quantified by densitometry on film as previously described (39).

The MIG6 3′-UTR reporter plasmid was constructed via insertion of the MIG6 3′-UTR (2561 bp) downstream of the Renilla luciferase stop codon in the pMIR vector (Promega) generating the pMIR-MIG63′UTR plasmid. For BIM, a commercially available 3′-UTR reporter plasmid, pEZX-BIM3UTR-1, was used (Genecopoeia). QuikChange Site-Directed Mutagenesis Kit (Stratagene) was used to generate mutations in the 3′ UTR of MIG6 and BIM by PCR using the pMIR-MIG6 3′UTR and pEZX-BIM 3′UTR as constructs templates. Primers containing the mutation TGCACTGA (1370–1377) → CCGGGCCG in the 3′UTR of MIG6 gene and TGCACTG (1029–1035) → GCGCGCC 3′UTR of BIM were used.

Glioblastoma cells were transfected with pre–miR-148a or pre-miR control for 6 hours. For MIG6, the cells were then transfected with either the reporter vector with 3′UTR-MIG6 or with mutant-3′UTR, in addition to a control β-galactosidase reporter plasmid. For BIM, the cells were transfected with either 3′UTR BIM or BIM-mutant-3′UTR. Luciferase assays were performed 48 hours later using the Luciferase System Kit (Promega) for MIG6 or the Dual Luciferase Assay (Promega) for BIM, and luminescence was measured on a Promega GloMax 20/20 luminometer. Firefly luciferase activity was double normalized by dividing each well first by β-galactosidase activity and then by average luciferase/β-galactosidase value in a parallel set done with a constitutive luciferase plasmid.

To determine whether MIG6 and BIM mediate the effects of miR-148a, rescue experiments were conducted in which the effects of anti–miR-148a were measured in the setting of inhibited MIG6 or BIM. Cells were either transfected with anti–miR-148a for 6 hours (1228) or U87 cells stably expressing anti–miR-148a were used. The cells were then transfected with siRNA against MIG6 (Thermo Fisher Scientific) or BIM (Cell Signaling Technology), and cell growth and death were assessed as described above. MIG6, EGFR, and BIM expression changes were verified by immunoblotting.

Cells were plated and transfected with either pre–miR-148a or pre-miR control for 24 hours followed by transfection with Rab7-mCherry for 24 hours (kindly provided by Marc G. Coppolino, University of Guelph, Ontario, Canada). Cells were serum starved overnight, followed by stimulation with 50 ng/mL EGF for 30 minutes. Samples were then washed, fixed, and permeabilized before immunostaining using primary antibodies (EGFR, Abcam; MIG6, Santa Cruz Biotechnologies). Samples were imaged using a 63× (NA 1.4) lens on a Zeiss LSM 700 with 405, 488, 543, 633 nm lasers using ZEN software (Carl Zeiss). Captured images were analyzed for colocalization using ImageJ software. Briefly, images were initially thresholded, and the Colocalization Finder tool was used to determine the area and intensity of colocalizing pixels of EGFR.

All experiments were performed at least three times. Two group comparisons were analyzed with t test and P values were calculated. For rescue experiments, the anti-miR-148a–induced change in the setting of inhibited target protein was compared with the anti-miR-148a–induced change in the control setting. For TCGA data, Cox regression analysis was performed to determine the correlation between miR-148a expression and patient survival. More detailed TCGA data statistical analyses are described in the corresponding sections. For all analyses, P < 0.05 was considered significant.

We analyzed TCGA data for miR-148a levels and for correlation with patient survival. The comparison of tumor (n = 491) with normal tissue samples (n = 10) showed a significant (59%) increase of miR-148a levels in the tumors as compared with normal brain (P = 3 × 10⁻⁴; Fig. 1A). Cox regression analysis of 482 glioblastoma samples in the TCGA dataset revealed that elevated miR-148a expression is a highly significant negative risk factor (P = 9.9 × 10-6). The HR was 1.19 with confidence intervals 1.10 to 1.29. The Kaplan–Meier curve of the TCGA patient cohort is shown in Fig. 1B. The lower quartile (with the lowest miR-148a expression) had longer overall survival than those with higher miR-148a expression. The median survivals of the different groups in the Kaplan–Meier curve are <25% expression = 515, 25%–50% = 463, 50%–75% = 377, 75%–100% = 382 (days). Log-rank analysis of 482 samples revealed that miR-148a was highly significant as a negative risk factor (P = 9.18 × 10⁻⁵; Fig. 1B). We also measured miR-148a levels in glioblastoma cells (U87, U373, T98G, A172, and SNB19), GSCs (0308, 0822, and 1228), and human tumor specimens (n = 18) as well as normal human astrocytes and normal brain (n = 7). MiR-148a was significantly higher in glioblastoma cells and GSCs than in astrocytes (P < 0.05; Fig. 1C) and significantly higher in tumors than in normal brain (P < 0.05; Fig. 1D). Altogether, these data demonstrate that miR-148a is upregulated in glioblastoma and that high miR-148a expression predicts poor patient survival.

We next assessed the functional role of miR-148a in glioblastoma (A172, SNB19, U87, and U373) and GSC (0308, 0822, and 1228) cells by determining the effects of miRNA overexpression and inhibition on cell growth and apoptosis using cell counting and Annexin V-7 AAD flow cytometry, respectively. miR-148a inhibition with antisense miRNA significantly decreased the growth rate (Fig. 2A) and overexpression of miR-148a resulted in a higher growth rate in glioblastoma and GSC cells as compared with controls (P < 0.05; Fig. 2B). Similarly, inhibition of miR-148a led to a significant induction of apoptosis (Fig. 2C), whereas overexpression of miR-148a led to a significant inhibition of apoptosis in glioblastoma cells and GSCs (P < 0.05; Fig. 2D). MiR-148a levels were verified by qRT-PCR (Supplementary Fig. S1). The above results show that miR-148a promotes cell growth and inhibits cell death in glioblastoma.

We next assessed the effects of miR-148a on glioblastoma cell migration and invasion. GSCs were not used for these experiments because they grow as neurospheres that do not attach to tissue culture plates. Anti–miR-148a or pre–miR-148a was transfected into glioblastoma cells followed by wound-healing and invasion assays. Inhibition of miR-148a expression significantly decreased (Fig. 3A) and overexpression of miR-148a significantly increased (Fig. 3B) the migration of glioblastoma cells. Inhibition of miR-148a expression significantly decreased (Fig. 3C) and overexpression of miR-148a significantly increased (Fig. 3D) the invasion of glioblastoma cells. These data show that miR-148a promotes glioblastoma cell migration and invasion.

We analyzed the effects of miR-148a on GSC self-renewal using a neurosphere formation assay. Anti–miR-148a or pre–miR-148a was transfected into GSCs and neurosphere formation was assessed for one week. MiR-148a inhibition significantly reduced neurosphere size and number and miR-148a overexpression increased neurosphere size and number (P < 0.05; Fig. 4A and B). These data suggest that miR-148a promotes the self-renewal ability of GSCs. To determine whether miR-148a affects GSC tumorigenesis, we assessed the effects of anti–miR-148a on orthotopic GSC xenograft formation. GSC 1228 cells were transfected with anti–miR-148a or anti–miR-control and stereotactic implanted into the striata of immunodeficient mice (n = 6). Tumor sizes were measured with MRI 4 weeks after implantation. Anti–miR-148a significantly inhibited tumor formation by GSCs (P < 0.05; Fig. 4C). We also assessed the effects of stable anti–miR-148a expression on glioblastoma xenograft growth. U87 cells stably expressing anti–miR-148a were orthotopically injected into nonobese diabetic/severe combined immunodeficient mice brains (n = 10) and tumor size was measured by MRI after 3 weeks. The result shows significantly reduced tumor volume in anti–miR-148a expressing xenografts as compared with controls (P < 0.05; Fig. 4D). These data show that miR-148a promotes GSC and glioblastoma tumor formation and growth.

To uncover mRNA targets of miR-148a in glioblastoma, we used bioinformatics databases (Targetscan, Pictar, RNAhybrid) to identify potential tumor suppressor targets. The following genes contained predicted binding sites for miR-148a: ERRFI1 (MIG6, NM_018948), BCL2L11 (BIM, NM_001204106), PTEN (NM_000314), SOCS3 (NM_003955), DNMT1 (NM_001130823), and JMY (NM_152405). To experimentally verify these potential targets, cells were transfected with miR-148a and assessed protein and mRNA target levels by immunoblotting and qRT-PCR, respectively. Two of the candidates were confirmed: MIG6 (ERRFI1) and BIM (BCL2L11). As MIG6 is a critical regulator of EGFR trafficking, degradation, and activation, we also determined the effects of miR-148a on EGFR expression and activation. MiR-148a inhibition increased (Fig. 5A) and miR-148a overexpression reduced (Fig. 5B) the expression of MIG6 in glioblastoma cells and GSCs. MiR-148a inhibition increased (Fig. 5C) and miR-148a overexpression reduced (Fig. 5D) the expression of BIM extralong (most abundant form of BIM) in glioblastoma cells and GSCs. Moreover, the effects of miR-148a on EGFR expression and activation were opposite to those on MIG6, as miR-148a overexpression led to increased EGFR and phospho-EGFR (Fig. 5B). We confirmed the above results in U87 cells stably expressing anti–miR-148 (Fig. 5E). MiR-148a also inhibited MIG6 and BIM mRNA levels, suggesting that its effects are via translation inhibition as well as via mRNA degradation (Supplementary Fig. S4). To determine whether MIG6 and BIM 3′-UTRs are direct targets of miR-148a, MIG6 or BIM 3′-UTR reporter constructs or 3′UTR-mutant controls were transfected into glioblastoma cells before transfection with miR-148a and luciferase activity was measured. Overexpression of miR-148a significantly reduced luciferase activity for both MIG6 and BIM (Fig. 5F). The above data show that miR-148a directly inhibits MIG6 and BIM and indirectly upregulates EGFR protein expression and promotes EGFR activation.

To determine whether the oncogenic effects of miR-148a are mediated by MIG6 and BIM, MIG6 or BIM upregulation by anti–miR-148a was prevented using siRNAs before assessment of cell growth (MIG6) or apoptosis (BIM). Glioblastoma cells were transfected with MIG6, BIM, or control siRNAs before transfection with anti–miR-148a followed by assessment of cell growth or apoptosis by cell counting and Annexin V-7AAD flow cytometry, respectively. Inhibition of miR-148a significantly inhibited glioblastoma and GSC cell growth. MIG6 knockdown partially prevented the effects of miR-148a inhibition on cell growth (Fig. 6A). Similar to earlier results, inhibition of miR-148a increased glioblastoma and GSC cell-line apoptosis; however, BIM knockdown prevented the increased apoptosis induced by anti–miR148a expression (Fig. 6B). MIG6 and BIM knockdown with siRNA was confirmed by immunoblotting (Fig. 6A and B). Similar rescue to the above was obtained in U87 cells stably expressing anti–miR-148a (Supplementary Figs. S5 and S6). The above data show that the oncogenic effects of miR-148a are partially mediated by MIG6 and BIM.

Previous research has shown that MIG6 regulates EGFR trafficking into the late endosome/lysosomes promoting EGFR degradation (11). We used confocal microscopy to determine whether miR-148a affects EGFR trafficking into a Rab7-positive late endosome/lysosomal compartment in glioblastoma cells. Rab7 has been shown to localize to late endosomes and to be important in the maintenance of the late endosomal compartment. Rab7 also controls the fusion of late endosome with lysosomes where EGFR degradation occurs (40). First, the glioblastoma cells were transfected with miR-148a or control before transfection with fluorescently labeled Rab7. We found reduced levels of MIG6 protein in miR-148a overexpressing cells as compared with control (data not shown). In control cells, MIG6 and EGFR colocalized in relatively large Rab7-labeled structures, likely multivesicular bodies (MVB)/late endosomes (Fig. 7A–O). This colocalization occurred at all time points, but was particularly evident 30 minutes after EGF stimulation in control cells (Fig. 7K–O, arrows). Importantly, in miR-148a–expressing cells, colocalization between EGFR, MIG6, and Rab7 was rarely seen and never found in the large Rab7-labeled structures (MVBs; Fig. 7P–T, grey circle). Colocalization is also shown in black and white for a clearer alternative image (Fig. 7E, J, O, and T). Quantification of the percentage of EGFR that colocalized with Rab7 and MIG6 showed a significant reduction in colocalization in miR-148a overexpressing cells compared with control cells (Fig. 7U). These data demonstrate that miR-148a reduces EGFR trafficking and degradation in glioblastoma cells.

MiR-148a has been investigated in some cancers but not in brain tumors (41–43). In this study, we investigated the expression, function, and mechanisms of action of miR-148a in glioblastoma. We found that miR-148a is a risk factor in glioblastoma where it acts as an oncogene by regulating BIM, MIG6, and EFGR stability and activation.

EGFR is one of the most frequently altered genes in glioblastoma. It is overexpressed in more than 60% of tumors but mutated and amplified in only about 40% (5, 44). Therefore, EGFR gene amplification only partially accounts for EGFR overexpression in glioblastoma (5), suggesting that additional mechanisms may be involved. Our study suggests that miR-148a overexpression is an important mechanism of EGFR overexpression via downregulation of MIG6. Consistent with our results, others have found that MIG6 expression is downregulated in approximately 50% of glioblastoma tumors without indications of MIG6 genomic deletions in the majority of samples (11). Our study also provides a new mechanism of MIG6 downregulation in glioblastoma.

We also identified the proapoptotic molecule BIM as a target of miR-148a, which is downregulated in 29% of glioblastoma cases based on TCGA analysis. Interestingly, a recent study demonstrated that elevated BIM expression levels in cancers strongly increased the antitumor activity of EGFR and other RTK inhibitors (45). These findings suggest that combined upregulation of BIM and inhibition of EGFR are likely to achieve synergistic antitumor effects. Our study shows that such combined targeting of BIM and EGFR can be achieved by inhibition of miR-148a, providing a rationale for the therapeutic targeting of miR-148a.

Previous research described miR-148a as a tumor suppressor in hepatocellular carcinoma, pancreatic cancer, gastric cancer, and colorectal cancer (42, 46–48). Our study demonstrates for the first time that miR-148a is oncogenic in glioblastoma. We show that miR-148a enhances glioblastoma and GSC growth, survival, migration, and invasion as well as GSC self-renewal and in vivo tumor growth. We also show that inhibiting miR-148a inhibits the above oncogenic endpoints. Importantly, based on our TCGA data analysis, we find that miR-148a expression displays a significant inverse correlation with glioblastoma patient survival. A recent study identifying a ten-miRNA prognostic expression signature in glioblastoma showed that miR-148a was among the seven miRNAs that were associated with high risk (49). Our TCGA data analysis expanded on this finding, analyzing 482 samples to further demonstrate elevated miR-148a expression in human glioblastoma specimens.

In summary, the present study shows that miR-148a is elevated in glioblastoma, where it predicts poor patient survival. It demonstrates that miR-148a has powerful oncogenic and cancer stem cell regulatory effects that are mediated by BIM, MIG6, and EGFR. The study therefore represents a first characterization of miR-148a as an oncogene and promising therapeutic target in glioblastoma.

No potential conflicts of interest were disclosed.

Conception and design: J. Kim, Y. Zhang, S. Lawler, R. Abounader

Development of methodology: B. Kefas, S. Parsons, S. Lawler

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D. Schiff

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Kim, Y. Zhang, M. Skalski, J. Hayes, S. Parsons, S. Lawler

Writing, review, and/or revision of the manuscript: J. Kim, M. Skalski, J. Hayes, B. Kefas, D. Schiff, B. Purow, S. Parsons, S. Lawler, R. Abounader

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Zhang, D. Schiff

Study supervision: Y. Zhang, S. Parsons, R. Abounader

Dedicated to the memory of Michael Skalski who passed away at a young age before seeing the fruits of his contribution to this work.

This work was supported by NIH grants NS045209 (R. Abounader) and CA134843 (R. Abounader).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.