The Histone Deacetylase Inhibitor Entinostat Mediates HER2 Downregulation in Gastric Cancer, Providing the Basis for Its Particular Efficacy in HER2-Amplified Tumors and in Combination Therapies

1. Cell culture

The cell lines MKN-7 (RRID:CVCL 1417), MKN-45 (RRID: CVCL_0434), MKN-74 (RRID:CVCL_2791), NCI-N87 (RRID: CVCL_1603), and OE33 (RRID:CVCL_0471) were obtained from the American Type Culture Collection (ATCC). Cells were cultured under standard conditions (37°C, 5% CO₂) in RPMI-1640 with phenol red and sodium pyruvate (Sigma Aldrich) supplemented with 10% fetal calf serum (SERANA) and 300 mg/L glutamine in a humidified incubator. Cell lines were authenticated by short repeat tandem profiling during the last 3years and cultured for less than 15 passages. All cell line cultures were regularly screened for mycoplasma contamination using a PCR Mycoplasma detection kit (Venor GeM Classic, minerva biolabs).

2. Cell treatment and transfection

Treatment of cells in vitro was performed 24 hours after seeding. Vehicle-treated cells (dimethyl sulfoxide [DMSO]) were used as negative control. siRNA was obtained from Eurofins MWG Operon (Ebersberg) and miRNA from Dharmacon. The sequence of HER2 siRNA was 5′-GCCUGAAUAUGUGAACCAGdTdT-3′ and 5′-CUGGUUCACAUAUUCAGGCdTdT-3′ for the sense and antisense strand, respectively. The sequence for the control siRNA (siLuc3) was 5′-CUUACGCUGAGUACUUCGAdTdT-3′ (sense strand) and 5′-UCGAAGUACUCAGCGUAAGdTdT-3′ (antisense strand). Transfection was carried out using INTERFERin (Polyplus), with 0.5 µL INTERFERin per pmol siRNA or miRNA according to the manufacturer’s protocol. For the transfections, 10 nM siRNA/miRNA were used.

3. RNA isolation and reverse transcription quantitative polymerase chain reaction

Depending on the cell line, 0.35-1.5×10⁵ cells were seeded per well of a 12-well plate. The RNA Magic Reagent (biobudget) was used for RNA extraction, according to the manufacturer’s protocol. The reverse transcription of the total RNA was performed using the RevertAid RT Kit (Thermo Fisher Scientific). The subsequent quantitative PCR was performed using the PerfeCTa SYBR Green FastMix ROX (QuantaBio) in the StepOnePlus Real-Time System (Thermo Scientific). The master mix was prepared according to the manufacturer’s protocol and the quantitative polymerase chain reaction was carried out under the following conditions: activation for 2 minutes at 95°C, followed by 45 cycles of 10 seconds at 95°C, 15 seconds at 55°C and 15 seconds at 72°C, with recording of the fluorescence intensity at the end of each cycle. For PCR product analysis, the samples were incubated at 65°C for 15 seconds and then heated up to 95°C to obtain a melting curve. For normalization, each sample was run with an actin-specific primer set and the target-specific primer set in parallel. Target levels were calculated by the formula 2^{(CP (actin)-(CP (target))}.

miRNA quantification was carried out using the qScript microRNA cDNA Synthesis Kit (Quantabio). Reverse transcription quantitative polymerase chain reaction (RT-qPCR) was performed according to the manufacturer’s protocol. The miRNA levels were calculated as above, by using a SNORD44-primer as a housekeeper for each sample. Primer sequences are given in S1 Table.

4. Protein isolation and immune blot

Depending on the cell line, 1-3×10⁵ cells were seeded per well of a 6-well plate. The next day, cells were treated for 72 hours, prior to harvesting and lysis in 20-50 µL RIPA buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% NP-40, 1% sodium deoxycholate) containing protease inhibitor cocktail set III, EDTA-free (1:500, Merck Darmstadt). Samples were centrifuged for 15 minutes at 12,000 ×g, prior to transferring the supernatant into a new reaction tube. Protein concentrations were measured using the BCA protein assay (Thermo Fisher Scientific) according to the manufacturer’s protocol and samples were stored at –80°C.

Protein samples (25-35 µg total protein) were mixed with loading buffer (125 mM Tris, pH 6.8, 20% glycerol, 4% sodium dodecyl sulphate [SDS], 2% β-mercaptoethanol, 10 µg/mL bromophenol blue) and heat-denatured for 5 minutes at 95°C, prior to separation by denaturing SDS-polyacrylamide gel electrophoresis. Protein bands were transferred from the 10% SDS-polyacrylamide gels by electroblotting onto a 0.2 µm nitrocellulose membrane. For saturation, the blots were incubated in 5% milk in TBST (Tris-buffered saline+Tween-20) for one hour, prior to primary antibody incubation overnight. Primary antibodies (for details, see S2 Table) were diluted in 5% milk in TBST to a concentration of 1:1,000 except for anti–glyceraldehyde 3-phosphate dehydrogenase (GAPDH) antibodies, which were used in a concentration of 1:5,000. After washing in TBST, blots were incubated with horseradish peroxidase secondary antibody (1:10,000) and, after an additional washing step, signals were detected by chemiluminescence (SuperSignal West Pico chemiluminescent substrate). Alternatively, fluorophore-coupled secondary antibodies, diluted 1:5,000 in 5% milk in TBST, were used and signals were detected at a wavelength of 800 nm in the same system. To quantify western blot results, bands on the same membrane (identical exposure conditions) were analyzed densitometrically using ImageJ (NIH). The densitometric values of the proteins of interest were then normalized to the values of a corresponding loading control.

5. Immunocytochemistry

For immunofluorescence, 1-3×10⁵ cells were seeded on glass plates in 6-wells, cultivated overnight and then treated with entinostat for 72 hours. After washing three times with phosphate buffered saline (PBS), cells were fixed with 2 mL Histofix (Carl Roth) per well. After three more washing steps, non-specific binding sites were blocked with 5% normal serum (from the same species as the secondary antibody) in PBS+0.3% Triton-X-100 for 1 hour. After three washing steps, the cells were incubated with the primary antibody at 4°C overnight. The next day, the cells were washed and the secondary antibody was added for 1 hour. After three more washing steps, the glass plates were transferred on slides and covered with DAPI (4’,6-diamidino-2-phenylindole)– containing mounting medium (Invitrogen).

6. Immunohistochemistry

From paraffin-embedded tumor pieces, 7 µm sections were prepared and baked at 56°C overnight, prior to microwave treatment for 5 minutes (800 W), deparaffinization in NeoClear (Carl Roth) and rehydration in a graded ethanol series. Epitope retrieval was performed in citrate buffer (29.4 g trisodium citrate dehydrate, 0.5% Tween-20 in 1 L dH₂O, pH 6.0) or EDTA buffer (0.37 g EDTA in 1 L dH₂O, pH 8.0) for HER2 or Ki-67 staining, respectively, at a temperature of 85-95°C for 15 minutes. After cooling for 10 minutes in the same buffer, the slides were washed three times in PBST (10 mM phosphate, 150 mM NaCl, 0.1% Tween-20) and incubated for 30 minutes in 2% H₂O₂ in PBST at 4°C for blocking endogenous peroxidases. After three more washing steps in PBST, nonspecific binding sites were blocked with 2% bovine serum albumin (BSA)/10% normal serum (from the same species as the secondary antibody) in PBST. Slides were then incubated at 4°C overnight with the primary antibody (S2 Table), prior to washing and addition of the biotinylated secondary antibody (1:500). After three more washing steps, slides were incubated with an avidin–horseradish peroxidase complex (VECTASTAIN ABC kit, Vector Laboratories) for 1 hour and washed again as above. The staining was developed by incubating the slides in a diaminobenzidine/H₂O₂ solution, prior to counterstaining with hematoxylin and mounting in Roti-Mount Aqua (Carl Roth).

7. Flow cytometry

Depending on the cell line, 1-3×10⁵ cells were seeded in 6-well plates, cultivated overnight and then treated with 1.0-5.0 µM entinostat (or DMSO 1:4,000 as negative control) for 72 hours. The cells were harvested by trypsinization and centrifuged. The cells were blocked by resuspending the cell pellet in 100 µL 1% Fc-block (BioLegend) in fluorescence-activated cell sorting (FACS) buffer (PBS+1% BSA) and incubating for 20 minutes at 4°C, prior to washing. Cells were then resuspended in 100 µL FACS buffer with anti-HER2 antibodies (1:200, coupled to fluorophore BV421, clone 24D2, BioLegend) and incubated for 20 minutes at room temperature. Another washing step was performed before the cells were resuspended in 100 µL 1% 7-aminoactinomycin D (BioLegend) in FACS buffer and put on ice until the measurement.

8. Annexin V/propidium iodide staining

To determine and quantitate viable, dead, necrotic, and apoptotic cells, a flow cytometry-based fluorescein isothiocyanate–annexin assay (Invitrogen) was performed according to the manufacturer’s protocol. For this, cells were seeded in 6-well plates as above and treated with entinostat for 48 hours. After treatment, the procedure was as described in the user guide, except for adding 395 µL binding buffer (10 mM HEPES, 140 mM NaCl, and 2.5 mM CaCl₂, pH 7.4)+5 µL propidium iodide (PI; 2 mg/mL) after the incubation period with annexin V. In order to be able to distinguish between the different apoptosis phases, PI (BL3-A) was gated against annexin (RL1-A). All events were detected per measurement.

9. Cell cycle analyses

Cell cycle analyses were performed after 48 hours entinostat treatment as above. The cells were trypsinized and fixed in 500 µL ice-cold ethanol (70%) for 30 minutes, washed twice, and incubated for 1 hour in 100 µL in FACS solution containing 50 ng/mL RNaseI (Attune Focusing Fluid, Thermo Scientific). Subsequently, 400 µL PI/FACS solution (2 mg/mL) was added. The cells were kept on ice until measurement in an Attune Acoustic Focusing Cytometer (Thermo Fisher Scientific).

10. Colony forming assay

Cells (1×10⁵ per well) were seeded in 6-well plates, treated with entinostat for 72 hours as described above, then trypsinized and counted in a Neubauer counting chamber (Carl Roth). Cells (1,000 per sample) were re-seeded in 6-wells in a total volume of 2 mL per well. After 2 weeks of growth, with a medium change every 72 hours, the colonies were stained with methylene blue (1 mg/mL in 50% (v/v) ethanol). Pictures were taken of the 6-well plates and the number of colonies was calculated using ImageJ (NIH).

11. Cell Counting Kit 8

Cells (1,000-5,000 per well) were seeded in 96-well plates and cultivated overnight, prior to treatment start (day 0). The number of viable cells was determined using the Cell Counting Kit 8 (CCK8; Dojindo) according to the manufacturer’s protocol. Briefly, the medium was removed and 50 µL CCK8 reagent, diluted 1:10 in medium, was added per well prior to incubation for 1 hour and measurement of the absorbance at 450 nm in an enzyme-linked immunosorbent assay plate reader, Multiskan FC (Thermo Scientific). As a blank value, the absorbance was measured in a well without cells and subtracted from the other values.

12. Subcutanous tumor xenograft models (cell line-based and patient-derived xenografts) and ex vivo tissue slice studies

Immunodeficient NOD/SCID/IL2r gamma (null) mice (Jackson Laboratory) were kept in a humidified atmosphere at 23°C, 12-hour light/dark cycle, with species-specific food and water ad libitum. Animal studies were performed according to the national regulations and approved by the local authorities (Landesdirektion Sachsen). All animal experiments complied with the ARRIVE guidelines and were carried out in accordance with the EU Directive 2010/63/EU for animal experiments.

To establish xenografts from cell lines, 5×10⁶ cells (MKN-74 or NCI-N87) in 150 µL PBS were subcutaneously injected into the flanks of the mice. The patient-derived xenograft (PDX) models were derived from gastric adenocarcinomas from patients treated in the Comprehensive Cancer Center Central Germany (CCCG, site Leipzig).

The tumor tissue was subcutaneously implanted into mice as described previously [11]. Briefly, mice were given metamizol intraperitoneally, followed by isoflurane anesthesia. A tumor specimen of 30 mm³ was implanted into a subcutaneous pouch in the flank region of the mice, prior to closing the wound with histoacryl tissue adhesive. The maximum time period between surgical removal of tumor tissue and primary engraftment of the mice was 4 hours. In order to prevent wound infection, mice received 1.45 mg/mL cotrimoxazol orally via drinking water for 10 days after transplantation. PDX tumor tissues were propagated in mice for at least three rounds.

When the xenografts reached a size of 6-8 mm in diameter, 350 µM thick tissue slices were prepared from the explanted tumors, using a vibratome. With a 3 mm biopsy punch, tissue slice pieces equal in size and excluding necrotic tumor areas were obtained and transferred onto cell culture inserts for air-liquid interface culture. After 24 hours, these tumor pieces were treated by removing the medium from the bottom compartment and replacing it with 1 mL medium containing 1.0-5.0 µM entinostat or DMSO (1:4,000 (v/v), as vehicle control). An untreated control was included as well. Tissue slice punches were harvested after 72 hours for RNA isolation or after 96 hours for the preparation of protein lysates or immunohistochemistry as described above.

For studying in vivo entinostat effects, PDX-bearing mice were randomized into three groups (three animals each) when the tumors reached a size of ~80 mm³. Mice were treated by intraperitoneal injection with low-dose or high-dose entinostat (1 mg/kg or 5 mg/kg body weight, respectively), or DMSO as negative control, on days 1, 3, 5, and 8, prior to termination of the experiment at day 9, surgical removal of the PDX tissue and further processing for RT-qPCR, western blotting and immunohistochemistry as described above.

13. Statistics

All experiments were performed for at least three times independently. Statistical significances were determined by one-way ANOVA/Holm-Sidak test using SigmaPlot14, and considered significantly different with p < 0.05, p < 0.03, and p < 0.01.

1. In vitro entinostat treatment reduces HER2 protein expression in HER2-amplified and non-amplified GC cells

Before evaluating entinostat effects on HER2 expression in GC cells, we screened a panel of human cell lines for basal HER2 expression. Western blot analyses of cell lysates (S3A Fig.) as well as flow cytometry of non-permeabilized cells for surface expression of the receptor (S3B Fig.) revealed major differences in basal HER2 levels. For the three cell lines with the highest HER2 expression, i.e., NCI-N87, MKN-7, and OE33 a genomic amplification of HER2 has been previously described [12-14]. In contrast, the cell lines MKN-74 and MKN-45 do not show any HER2 amplification [12] as reflected here by their lower HER2 protein expression (S3A and S3B Fig.). Of note, MKN-45 cells showed a particularly low surface expression of HER2 (S3B Fig.) despite somewhat higher expression in whole cell lysates (S3A Fig.). The selection of cell lines with very high or rather low HER2 levels, respectively, for subsequent experiments also allowed us to evaluate whether entinostat effects on HER2 expression and/or cell proliferation are dependent on the basal expression of this receptor or the presence/absence of HER2 gene amplification.

On the mRNA level, entinostat treatment led to no major alterations in HER2-amplifed MKN-7 and NCI-N87 cells (Fig. 1A). In HER2 non-amplified MKN-45 cells a mild ~2-fold induction occurred (Fig. 1A), while in HER2 non-amplified MKN-74 cells a HER2 mRNA downregulation was found (Fig. 1A).

Fig. 1.

Effect of entinostat treatment vs. dimethyl sulfoxide vehicle as designated by D on human epidermal growth factor receptor 2 (HER2) mRNA levels (48-hour treatment) (A) and HER2 protein expression (72-hour treatment) (B) in NCI-N87, MKN-7, MKN-74, and MKN-45 cells as determined by reverse transcription quantitative polymerase chain reaction using a HER2-specific primer pair (A) and western blotting using a monoclonal antibody directed against the C-terminus of HER2 (B). Bar diagrams: quantitation of band intensities (mean±standard error of mean [SEM], n=3); lower panel: representative immunoblots for HER2 and glyceraldehyde 3-phosphate dehydrogenase (GAPDH) as loading control. ^a)p < 0.05, ^b)p < 0.03, ^c)p < 0.01.

In contrast, entinostat caused a substantial, dose-dependent downregulation of HER2 protein levels in all four cell lines (Fig. 1B). For analyzing this effect, a HER2-specific antibody recognizing an epitope at the C-terminal site of the receptor was employed (3B5) (S2 Table). Notably, a pronounced downregulation was also found in the non-amplified cell lines MKN-45 and MKN-74 (Fig. 1B) demonstrating that the entinostat effects on HER2 expression were not restricted to cell lines with high expression of HER2. In this context, it should be noted that the HER2 bands were recorded under optimized exposure conditions, for optimal comparison of the band intensities within a given blot. Thus, the HER2 band intensities between the different blots do not allow for comparing basal expression levels of the receptor in the different cell lines; see supplementary data for a direct comparison (S3 Fig.). To further corroborate our findings, we also used a different antibody directed against the N-terminus of the receptor (N12) and obtained fully comparable results, as shown here in the cell lines OE33 and NCI-N87 as representative examples (S4A Fig.). The HER2 downregulation was also seen in flow cytometry using a third anti-HER2 antibody (24D2) in non-permeabilzed cells (S4B Fig.). Thus, we conclude that entinostat markedly reduces total and cell surface HER2 protein expression and that this effect is independent of the basal HER2 levels (before HDACi treatment) or the presence/absence of a HER2 gene amplification. Moreover, the fact that the downregulation of HER2 was seen using antibodies recognizing the C or N-terminus of HER2 suggests that this effect is not based on posttranslational cleavage or shedding of the receptor.

2. HER2 protein is downregulated by entinostat in ex vivo slice cultures of GC xenografts

Since conventional, two-dimensional (2D) cell culture on a plastic substrate does not necessarily reflect the situation in the tumor tissue in vivo, we extended our experiments towards three-dimensional (3D) tumor slice cultures, for studying intact tumor tissue ex vivo. For this, tumorigenic MKN-74 (no HER2 amplification) or NCI-N87 (HER2-amplified) cells were subcutaneously injected into immunocompromised mice, and after 4-5 weeks mice were sacrificed and tumor xenografts excised for ex vivo tissue slice preparation and culture (see Material and Methods, for further details). In NCI-N87 xenograft tissue slices with high HER2 expression, entinostat effects (Fig. 2A) were found to be almost identical to the findings in conventional in vitro 2D cell culture (Fig. 1A and B, right). Again, no relevant effects were seen on the mRNA level, whereas profound HER2 downregulation was observed on protein level (Fig. 2A). Notably, in the case of MKN-74 cells with very low basal HER2 expression levels (S3 Fig.), the inhibitory entinostat effects on HER2 expression were even more pronounced in the 3D tissue slice culture (Fig. 2B) as compared to the findings in conventional 2D cell culture (Fig. 1A and B, second panel from right). The entinostat-mediated HER2 downregulation was also associated with reduced proliferation, as determined by immunohistochemical staining for the proliferation marker Ki67 (Fig. 2C).

Fig. 2.

Direct comparison of entinostat-mediated human epidermal growth factor receptor 2 (HER2) downregulation in NCI-N87 cells with high basal HER2 expression (A) and MKN-74 with low basal HER2 expression (B). For both cell lines mRNA expression of HER2 (A, B, upper) and protein expression (A, B, lower) were analyzed in ex vivo tissue slice models. Lower panels in (A, B, lower): representative examples of immunoblots for HER2 and glyceraldehyde 3-phosphate dehydrogenase (GAPDH) as loading control; upper panels (bar diagrams) in (A, B, lower): quantification of the western blot band intensities (mean±standard error of mean [SEM]) of three independent experiments. ^a)p < 0.01. Cells were treated for 48 hours for mRNA analyses and 72 hours for protein analyses. In addition, proliferation inhibition in NCI-N87 tumor xenograft tissue slice cultures is shown, as determined by reduced expression of the proliferation marker Ki-67 (C). Untreated slices vs. vehicle control dimethyl sulfoxide (DMSO) vs. entinostat 1 or 3 µM treatment for 72 hours are presented. Expression levels of Ki-67 were determined by immunohistochemistry (brown; see red arrowheads). Sections were counterstained with hemotoxylin.

In summary, in most cell lines the downregulation of HER2 protein expression could not be explained by transcriptional effects on mRNA levels, since only in MKN-74 cells a decrease in HER2 mRNA levels occurred in 2D cell culture and slice culture models. Rather, the downregulation of HER2 by entinostat is identified as a post-transcriptional effect.

3. Downregulation of HER2 by entinostat is associated with impairment of AKT signaling, whereas entinostatpromoted p21 induction is independent of effects on HER2

The consistent downregulation of HER2 protein expression upon entinostat treatment found in different assays and different cell systems suggests an impairment downstream signaling pathways. However, the inhibitory effect of entinostat on HER2 may be counterbalanced by adaptive induction of compensatory alternative (HER2-independent) pathways, preventing the inhibition of downstream signal transduction pathways. Thus, effects of entinostat were analyzed regarding phosphorylation of AKT as a canonical effector of receptor tyrosine kinase (RTK)–mediated proliferative responses. For comparison, unphosphorylated AKT and GAPDH, as an AKT-independent loading control, were analyzed in parallel in the western blots. Moreover, effects on p21 expression were monitored as a possible target whose upregulation may indicate tumor-inhibitory effects.

In both NCI-N87 and MKN-74 cells (with or without HER2 amplification, respectively) a dose-dependent p21 induction was found, suggesting a general tumor-inhibitory effect (Fig. 3A and B). Notably, this was seen in 2D cell culture as well as in the 3D tumor xenograft tissue slice models derived from the same cell lines. 2D cell culture was found to be slightly more sensitive towards entinostat, as indicated by the onset of the p21 upregulation already at lower entinostat concentrations (compare MKN-74, 1 µM), the occurrence of dose-limiting entinostat cytotoxicity in MKN-74 cells at 5.0 µM entinostat in 2D, but not 3D culture, and the lower plateau of p21 levels in the NCI-N87 tissue slices. While these differences may be based on decreased drug penetration in the intact 3D tumor tissue, both systems revealed overall very comparable entinostat effects on p21.

Fig. 3.

Effects of entinostat treatment (72 hours) on the cell cycle regulator p21 (A, B) and the human epidermal growth factor receptor 2 downstream target p-AKT (C, D) in the cell lines MKN-74 and NCI-N87, in 2D cell culture and in the corresponding tumor xenograft tissue slice cultures. For comparison untreated and dimethyl sulfoxide vehicle–treated samples are shown as designated by U and D, respectively. Lower panels show representative immunoblots in (A, B) for p21 with glyceraldehyde 3-phosphate dehydrogenase (GAPDH) as loading control and in (C, D) for p-AKT with unphosphorylated AKT and GAPDH as loading controls; bar diagrams: quantitation of band intensities (mean±standard error of mean [SEM]) of three independent experiments. ^a)p < 0.05, ^b)p < 0.03, ^c)p < 0.01.

In contrast, an entinostat-mediated marked inhibition of AKT phosphorylation in MKN-74 cells was only observed in conventional 2D cell culture, but not in 3D tissue slices from MKN-74 xenografts (Fig. 3C). The unexpected increase in AKT phosphorylation in the more complex tissue slice model suggests that adaptive responses occur in some cell lines after entinostat treatment. The difference between 2D and 3D cell culture systems could indicate the involvement of stromal factors, such as soluble HER ligands, and thus provide a starting point for further studies. In the HER2-amplified NCI-N87 cells, however, the decrease in phosphorylated AKT was observed in both systems, with very comparable results (Fig. 3D). These findings thus highlight profound molecular alterations upon entinostat treatment as well as the relevance of ex vivo tissue slice culture systems in parallel to classical 2D cell culture for monitoring molecular responses (activation or inhibition of signaling pathways).

To evaluate whether the observed downregulation of HER2 was necessary and sufficient for explaining the effects on AKT phosphorylation in HER2-amplified cells, results were compared to an siRNA-mediated knockdown in NCI-N87 cells. Indeed, a ~50% knockdown of HER2 (S3A and S3B Fig.) already led to a substantial reduction of phosphorylated AKT (S5A and S5C Fig.). While this was in line with the above entinostat effects, diverging results were obtained regarding p21 levels, i.e., HER2 siRNA led to a downregulation of p21 (S5A and S5D Fig.). These findings support the notion that the p21 effects of entinostat are independent of inhibiting HER2 expression, while the inhibition of AKT signaling, at least in HER2-amplified cells, is closely associated with the downregulation of HER2.

4. Cell line-specific contribution of HER2-inhibiting microRNA miR-205 to entinostat effects

In the next step, we investigated possible effects of entinostat on the levels of HER2-regulatory microRNAs (Fig. 4A). In fact, beyond modulating gene regulation on the level of transcription or regulating mRNA stability, miRNAs are also able to repress translation [15-17].

Fig. 4.

Entinostat effect on microRNA levels and microRNA-dependent regulation of human epidermal growth factor receptor 2 (HER2) expression. (A) Schematic illustration of miRNAs regulating HER2 and/or affected by HER2. (B) Alterations in miRNA expression levels in NCI-N87 cells after entinostat treatment for 48 hours with the indicated concentrations. (C) Time-dependent reduction of HER2 protein levels upon transfection of NCI-N87 cells with 10 nM of miR-205-3p. (D) Effect of anti-miR transfection on HER2 expression upon entinostat treatment. HER2 levels after transfection of control miRNA (ctrl) were compared with HER2 levels after transfection of specific miR-205 inhibitor (205). For the protein analyses HER2 levels were monitored 72 hours after miR (C) or anti-miR (D) transfection. Mean±standard error of mean (SEM) of three independent experiments are given. ^a)p < 0.05.

Based on in silico predictions and the literature, a number of miRNAs have been identified to directly affect the expression of HER2 or its modulation by inhibitors, or are themselves expressed in a HER2-dependent manner [10]. For example, miR-125 has been described as direct regulator of HER2 and modulator of anti-HER2 therapy response. However, entinostat treatment of NCI-N87 cells did not alter miR-125 levels, thus excluding this miRNA as mediator of entinostat effects (Fig. 4B). While the same was true for miR-331-3p, entinostat treatment led to a very profound upregulation of miR-205 (Fig. 4B).

Interestingly, miR-205 is not only an upstream regulator of HER2, but vice versa, it is also regulated by HER2 signaling (Fig. 4A) [10]. Thus, the observed induction of miR-205 could also represent just a consequence of HER2 downregulation. Therefore, we used miR-205 transfection experiments to test whether miR-205 could mediate a HER2 inhibition in GC cells. In fact, miR-205 transfection led to a profound and sustained inhibition of HER2 protein levels in NCI-N87 cells. This was seen for both, miR-205-3p and miR-205-5p, indicating its functional relevance as mediator of the inhibition of HER2 by entinostat (Fig. 4C). Likewise, miR-205 transfection also decreased HER2 protein expression in HER2 low expressing MKN-74 cells (S6A Fig., left) suggesting that this effect was not independent of the basal HER2 expression. However, in clear contrast to the findings in NCI-N87 and MKN-74 cells, miR-205 transfection did not affect HER2 levels in HER2-amplified MKN-7 cells (S6B Fig., left) and in non-amplified MKN-45 cells (S6C Fig., left). This suggests cell-context–dependent differences in the mechanism of HER2 regulation and the possibility that miR-205–independent effects are involved as well.

The influence of a miR-205–specific anti-miR on the effect of entinostat was tested next. Here, cell line–dependent effects were observed as well: in NCI-N87 (Fig. 4D) and in MKN-45 cells (S4C Fig.), miR-205 inhibition led to higher HER2 levels under entinostat treatment, while no such effect was detectable in MKN-74 (S6A Fig.) and MKN-7 cells (S6B Fig.). Thus, miR-205 effects were found to be dependent on the cell line.

5. HER2-downregulation and impairment of AKT signaling by entinostat is also observed in ex vivo cultivated PDX

So far, our experiments had shown consistent effects of entinostat on HER2 protein expression in various GC cell line-based in vitro (2D culture) and ex vivo (3D tumor xenograft tissue slice) model systems, with, however, some differences in the mechanism of HER2 downregulation. To test whether our findings are also reproducible in more heterogenous patient tumor material, studies were extended towards ex vivo slice cultures from PDX. Primary tumors from two patients were propagated in immunocompromised mice and led to the establishment of stable PDX tumor models. The PDX tumors essentially retained the initial tissue morphology, tumor heterogeneity, and HER2 expression levels of the primary tumor (not shown). Thus, PDX models reflect the clinical situation more accurately as compared to tumor cell lines, which have undergone selection and adaptation processes to 2D cell culture conditions upon continuous growth in vitro over very long time periods. It should be also noted that the pathological analysis of the primary tumors revealed the tumor of patient 1 to belong to the MSI subtype of GC. This further extended the spectrum beyond our microsatellite stable cell line models tested so far.

The analysis of HER2 mRNA levels in the PDX tissue slices upon entinostat treatment largely confirmed the previous data in the cell lines. While no major alteration in HER2 mRNA expression was observed in the patient 1 PDX (Fig. 5A), a slight dose-dependent trend towards decreased HER2 mRNA levels was found in the PDX tissue of patient 2 (Fig. 5C). Notably, in both PDX models a very profound downregulation of HER2 was detected on the protein level (Fig. 5B and D, left panels), with HER2 bands in western blot experiments being close to, or even below, the limit of detection at higher entinostat concentrations. This was again associated with a marked and dose-dependent inhibition of AKT phosphorylation (Fig. 5B and D, center panels) and with the induction of p21 (Fig. 5B and D, right) in both patient models. Still, some minor differences between the two models were observed, with p-AKT downregulation being more prominent in the patient 1 PDX tissue while patient 2 showed a particularly pronounced upregulation of p21.

Fig. 5.

Entinostat effects on human epidermal growth factor receptor 2 (HER2) expression (mRNA and protein) and downstream effectors (phospho-AKT and p21) in tissue slice cultures from patient-derived xenograft (PDX) tumors. (A, B) Patient 1 PDX model. (C, D) Patient 2 PDX model. After 72 hours of entinostat treatment as detailed in the figures the tissues were analyzed for HER2 mRNA levels (A, C) or HER2, phospho-AKT, and p21 protein levels (B, D) as described above. Mean±standard error of mean (SEM) are given. ^a)p < 0.05, ^b)p < 0.03, ^c)p < 0.01.

6. In vivo entinostat treatment increases miR-205 and reduces HER2 expression in PDX tumor tissue

Pharmacological intervention in vitro and ex vivo does not fully reflect pharmacokinetic aspects such as drug uptake, distribution, metabolism, excretion, and tissue penetration, leaving it questionable whether the drug concentrations investigated in cell culture are realistically achievable at their site of action in vivo. For this reason, we next tested an in vivo intervention in PDX tumor-bearing mice by intraperitoneal application of entinostat.

Tumor PDX samples from patient 1 (see above) were used in this experiment. Subcutaneous implantation of tumor tissue pieces with approximately 1-2 mm in diameter into NSG mice led to the establishment of tumor nodules within 4 weeks, with a tumor take rate of 90%. When tumor sizes were ~80 mm³, mice were randomized into three groups for studying entinostat effects in vivo: control group (DMSO), entinostat low-dose and entinostat high-dose group (1 mg/kg and 5 mg/kg body weight, respectively). Upon termination of the experiment after 4 consecutive treatments over 8 days, the RT-qPCR–based analysis of tissue lysates revealed no differences in HER2 mRNA levels (Fig. 6A). While this was in line with the previous in vitro and ex vivo data, it should be noted that this was not based on poor drug delivery to its site of action. Rather, the analysis of miR-205-3p levels revealed a marked ~4-fold upregulation upon entinostat treatment (Fig. 6B). Likewise, entinostat effects on HER2 protein levels were also observed in vivo, with a dose-dependent reduction of HER2 protein in tissue lysates to ~60% (Fig. 6C). This reduction in HER2 levels was also confirmed by immunohistochemistry, clearly establishing in vivo the entinostat-mediated HER2 downregulation (Fig. 6E). To test the specificity of the antibody for IHC, we included control experiments with NCI-N87 cells in 2D cell culture via immunocytochemistry (S7A Fig.) and in tumor xenografts via immunohistochemistry (S7B Fig.) proving the membrane staining by the HER2 antibody. In addition, to these short-term in vivo experiments, we also performed longer-term experiments over a period of 19 days to delineate potential growth-inhibiting effects of the entinostat therapy. In fact, the HDAC inhibitor proved a dose-dependent inhibition of tumor growth with a 40% reduction in tumor volume after application of the higher entinostat dose (Fig. 6D).

Fig. 6.

In vivo entinostat effects on human epidermal growth factor receptor 2 (HER2) levels in patient-derived xenograft (PDX)–bearing mice. After establishment of PDX tumors, the mice (n=3 per group with two tumors per mouse) were treated for a short-term experiment with entinostat in a dose of 1 mg/kg or 5 mg/kg body weight per injection, respectively) or dimethyl sulfoxide (DMSO) as negative control. The animals were intraperitoneally injected four times within 7 days. Tumors (n=6 per group) were analyzed on day 8 for HER2 mRNA levels (A) and miR-205-3p expression (B) via reverse transcription quantitative polymerase chain reaction and HER2 protein levels, as determined by Western blotting (C) and immunohistochemistry (E). In addition, a long-term experiment was performed to measure the effect of entinostat on tumor growth. For this purpose, 5-6 mice per group were treated by intraperitoneal injection three times per week for 19 days. The tumor growth rate is shown in (D) for the different treatment groups with entinostat in a dose of 1 mg/kg or 5 mg/kg body weight, respectively. Mean±standard error of mean (SEM) are given. ^a)p < 0.03, ^b)p < 0.01.

7. HER2-amplified GC cells are more susceptible towards entinostat treatment than non-amplified cells

It is well established that HER2 (over-)expressing GC cells are dependent on intact signaling via this receptor and therefore show pronounced growth inhibition/apoptosis upon HER2 inhibition. Therefore, we hypothesized that entinostat, through its HER2 downregulating effect, would more severely impair the growth of HER2-overexpressing cells compared with non-overexpressing cells. Indeed, CCK8-based proliferation assays upon treatment with 1 µM entinostat revealed a more profound inhibition of NCI-N87 cells with a HER2 gene amplification as compared to MKN-74 cells (> 80% vs. ~50%) (Fig. 7A). At 3 µM entinostat, an essentially complete inhibition was seen in both cell lines, suggesting the expected additional (HER2-independent) entinostat effects at this higher concentration.

Fig. 7.

Cellular effects of entinostat treatment on proliferation or cell death. (A) Effect of treatment with entinostat (1 or 3 µM) vs. untreated or vehicle (dimethyl sulfoxide [DMSO]) treated cells was evaluated by Cell Counting Kit 8 (CCK8) assay in MKN-74 (low basal human epidermal growth factor receptor 2 [HER2] expression) or NCI-N87 (high basal HER2 expression) cells. Treatment started at day 0 and no medium change was performed. (B) Evaluation of colony formation in MKN-74 and MKN-45 cells (both with low basal HER2 expression and MKN-7, OE-33, and NCI-N87 cells (high basal HER2 expression). Note that the cells were treated with entinostat (1 or 3 µM) only for 72 hours (see ”Colony Forming Assay” in Materials and Methods) before evaluation colony growth over 2 weeks without any further treatment, which explains the lower inhibitory effect in MKN-74 cells in comparison to the CCK8 assay with a continuous presence of the inhibitor. Mean±standard error of mean (SEM) are given. (C) Cell cycle (flow cytometry–based counting of propidium iodide [PI] stained cells), and (D) cell death analyses (flow cytometry–based counting of PI/annexin stained cells) in MKN-74 and NCI-N87 cells. Cells were treated with entinostat for 48 hours with the indicated concentrations of entinostat or with DMSO as vehicle control (designated as D) or left untreated (designated as U).

The HER2-dependency of entinostat at lower concentrations was even more evident in colony formation assays (Fig. 7B). While 1 µM entinostat led to only minor inhibition of colony formation in both cell lines with low HER2 expression (MKN-74 and MKN-45; blue arrows), the same treatment resulted in substantial reduction of colonies in the HER2-amplified cell lines (red arrows). These findings indicate that high expression of HER2 is associated with a higher sensitivity towards entinostat.

Effects of entinostat treatment were also seen on cell cycle. PI staining and FACS analysis revealed a moderate reduction of cells in G2/M upon treatment with 3 µM entinostat (Fig. 7C). This effect was slightly more profound in HER2 overexpressing NCI-N87 cells as compared to MKN-74 cells. The proportion of cells in S phase was found to be diminished as well, while an increase in a fraction with reduced DNA content, corresponding to sub-G1 cells, was observed. Interestingly, these findings were somewhat more profound in MKN-74 cells (Fig. 7C). The latter finding is in line with the well-known DNA damaging activity of HDAC inhibitors, which relies on HER2-independent effects like impairment of DNA repair or induction of reactive oxygen species. In general, these results were also found in MKN-45 and MKN-7 cells, with both cell lines exhibiting an inhibition in the G2/M phase (S8A and S8B Fig., right panel) and with MKN-45 cells showing a significant inhibition in the S phase (S8A Fig., center panel).

Concomitantly, FACS-guided PI/annexin V staining for detecting entinostat-mediated induction of cell death revealed an increase in early (annexin V–positive, PI-negative) and late apoptotic cells upon treatment with 3 µM entinostat (annexin V–positive, PI-positive cells) (Fig. 7D). Here again, a higher susceptibility of NCI-N87 cells in comparison to MKN-74 cells was observed. Thus, cell death induction complements the above findings on cell proliferation (Fig. 7A) or clonogenic cell survival (Fig. 7B) and supports the notion that higher HER2 expression contributes to higher entinostat sensitivity. Interestingly, MKN-74 cells, which exhibited a more pronounced sub-G0 fraction after entinostat treatment compared to NCI-N87 (Fig. 7C), showed lower stainability for annexin V/PI (Fig. 7D). This proves the non-redundancy of these two read-outs, for example in the sense that upon entry into apoptosis in cell cycle phases other than G0/G1, no corresponding and analogous sub-G0 peak may occur despite an increase in annexin V/PI-positive cells (for overview [18]). In addition, signals in the subG0 fraction may also reflect debris from cells that have died via alternative cell death mechanisms as well as aneuploid cells. Of note, clear pro-apoptotic effects of entinostat were also observed inr MKN-45 and MKN-7 cells (S9A and S9B Fig., lower panels), with MKN-45 cells showing a particular increase in late apoptotic cells (PI +/Ann +), whereas MKN-7 cells demonstrating particularly pronounced increase in early apoptotic cells (PI –/Ann +).

8. Entinostat plus HER2 kinase inhibitor as a useful combination therapy in low HER2-expressing GC cells

Finally, we tested the efficacy of dual treatment of GC cells with entinostat and the HER2-specific kinase inhibitor CP724714. The rationale for this combination treatment was based on previous findings that treatment of cancer cells with HER2 kinase inhibitors can lead to a rapid adaptive increase in HER2 expression [19,20]. Given the fact that in our hands entinostat showed higher antineoplastic potency in GC cells with higher HER2 expression (Fig. 7A, C, and D) we hypothesized that an adaptive HER2 induction upon kinase inhibitor treatment would lead to higher entinostat susceptibility. Thus, we assumed that entinostat would add to the antitumor effects of HER2 kinase inhibition.

In fact, treatment with 2.0 µM of the HER2 kinase inhibitor CP724714 led to a ~2-fold induction of HER2 expression in MKN-74 cells. While single treatment with 1.5 µM entinostat was not able to reduce HER2 levels, indicating that this concentration was too low in this cell line as seen before (Fig. 2B, left), it was able to fully revert the CP724714-mediated induction upon combination of both inhibitors (Fig. 8A). More importantly, the combination of entinostat and CP724714 proved to be clearly more inhibitory on proliferation of MKN-74 cells than any of the single substances (Fig. 8B). Specifically, the HER2 inhibitor, which had no antiproliferative effect as a single substance (in line with the rather low basal HER2 expression of MKN-74 cells), caused an increase in entinostat toxicity in the combination treatment (Fig. 8B). Both, the non-effect of CP724714 alone and the favorable combination effect on proliferation may be explained by the upregulation of HER2 associated with HER2 inhibitor treatment, which could lead to an acquired vulnerability against entinostat. In this context, it should be noted that the effects of the HER2 inhibitor are ambiguous in that HER2 induction (i.e., an undesirable, potential pro-oncogenic effect) may be partially counter-acted by an inhibition of HER2 kinase activity. Consequently, the resulting cumulative effect may be difficult to predict.

Fig. 8.

Effect of combination therapy with entinostat plus selective human epidermal growth factor receptor 2 (HER2) inhibitor CP724714. (A) Analysis of HER2 protein expression in MKN-74 cells treated for 72 hours with vehicle dimethyl sulfoxide (DMSO, D), 1.5 μM entinostat (E), 2.0 μM CP-724714 (CP), or the combination of both (E+CP). Densitometric quantifications of three independent experiments (mean±standard error of mean [SEM]) and one representative western blot are shown. (B) Effect of entinostat, CP724714, or the combination of both on proliferation of MKN-74 cells in comparison to vehicle-treated cells as determined by Cell Counting Kit 8 (CCK8) assay (mean±SEM, n=3). (C) Effect on colony formation of MKN-74 after treatment with entinostat, CP, or entinostat+CP. Scale bars=500 µm. (D) Combinatory effect on cell death, analysis was conducted using flow cytometry-based distribution of propidium iodide (PI) and annexin-V–stained cells. (E) Combinatory effect on cell cycle, analysis was conducted using flow cytometry–based cell cycle distribution of PI-stained cells. Quantitation of three independent experiments (mean±SEM) are shown in (D) and (E). ^a)p < 0.05, ^b)p < 0.03, ^c)p < 0.01.

To further corroborate combination effects of entinostat and HER2 inhibitor, additional colony-forming assays (Fig. 8C, S8 Fig.) as well as apoptosis assays (Fig. 8D, S10 Fig.) and cell cycle analyses (Fig. 8E, S11 Fig.) were performed. To summarize these data, we found that cell lines with high HER2 expression (e.g., NCI-N87 or MKN-7) already showed strong inhibition under single treatment with entinostat (NCI-N87 and MKN-7) or even the HER2 inhibitor CP274714 (NCI-N87), which diminished possible additional effects of a combination therapy. In contrast, cells with lower HER2 expression (MKN-74 or MKN-45) were clearly less susceptible to either inhibitor alone, but showed stronger effects of the combination therapy.